University of Groningen

FACULTY OF SCIENCEAND ENGINEERING

Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence

ANNUAL REPORT 2018

Bernoulli Institute Annual Report

Introduction

This is the annual scientific report over 2018 of the Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence, which was established on the 1st of June 2018 when the former Johann Bernoulli Institute for Mathematics & Computer Science (JBI) and the Institute for Artificial Intelligence & Cognitive Engineering (ALICE) joined forces. The mission of the Bernoulli Institute is to perform outstanding academic research and teaching in Mathematics, Computer Science, and Artificial Intelligence, and to maintain international leadership herein; to foster these disciplines as a living body of knowledge, and to make it relevant to society in its broadest sense. The symbiosis between pure and applied science, and between mono- and multidisciplinary research and teaching, is a distinguishing characteristic of our institute. As an important part of this mission we aim to transfer our results to other areas of science and technology, and initiate and expand inter- and multi-disciplinary research collaborations. Within the Faculty of Science and Engineering the institute has a leading role in the cross- disciplinary research Center for Data Science and Systems Complexity (DSSC), and in the Groningen Cognitive Systems and Materials (CogniGron) Center, which is a joint enterprise of the Bernoulli Institute and the Zernike Institute for Advanced Materials. A total of twelve professor positions have been opened within CogniGron, ten of which have been designated to the Bernoulli Institute. In July of this year the Minister of Education, Culture and Science decided about the division of the Sector Plan budget over the Dutch universities. Four positions in Mathematics and six positions in Computer Science/Artificial intelligence have been allocated for the Bernoulli Institute. An initial call for these positions took place in 2018 in the context of the Rosalind Franklin programme of the University of Groningen. As a result of the above developments a strong growth of the institute is foreseen in the coming years.

Some statistics

In 2018 the institute had 53 (tenured and tenure track) scientific staff members and 9 support staff members. A total of 116 PhD candidates were enrolled, including 4 Ubbo Emmius scholarship students from abroad, 10 PhD students funded by the Netherlands Organisation for Scientific Research (NWO) and 72 PhD students funded by the European Union, industry, or other external funding. Also 12 postdocs worked at the institute.

1 Bernoulli Institute Annual Report

A total of 19 doctoral dissertations were successfully defended. A total of 175 journal papers, 27 (contributions to) books, 120 refereed contributions to conference proceedings and 13 other professional publications were published. Members of the institute served as editors-in-chief, associated editors or members of the editorial boards of international journals and book series. The institute was visited by almost hundred scientists from abroad.

Organisation

Personalia

Dr. Dimka Karastoyanova was appointed as full professor of Information Systems and head of the Information Systems Group at the Department of Computer Science in January 2018. Dimka Karastoyanova had a joint appointment by the KLU in Hamburg and HPI of the University of Potsdam as an Associate Professor of Data Science and Business Intelligence and Senior Researcher, respectively, from September 2016 to December 2017. She was a junior professor in Simulation Workflows at the University of Stuttgart in the scope of Cluster of Excellence SimTech (Simulation Technology) from November 2008 to August 2016. She received her doctoral degree in Computer Science from the Technische Universitat¨ Darmstadt in 2006. Her research is in the broader field of information systems and more specifically in application integration, middleware, Service Oriented Architecture, Business Process and Workflow Management with special focus on adaptive systems, Scientific Workflows, Provenance, Distributed and Adaptable Scientific workflow management systems and their security aspects. The application fields range from conventional business applications, through scientific computing applications, to healthcare and logistics. Dr. Bart Besselink has been appointed as tenure-track assistant professor on the CogniGron position Engineering Mathematics. He was embedded in the Systems, Control and Applied Analysis group in August 2018, after holding an assistant professor position in the same group since August 2016. He obtained his PhD from Eindhoven University of Technology in 2012 and was a postdoctoral researcher at KTH Royal Institute of Technology, Stockholm, Sweden, between 2012 and 2016. His research interests are in the analysis and control of large-scale interconnected systems with emphasis on the problems of compositional analysis and model reduction. Dr. Marcello Seri has been appointed as tenure-track assistant professor within the GQT Mathematics cluster. He was embedded in the Dynamical Systems, Geometry & Mathematical Physics Group in June 2018. His research interests revolve around mathematical physics problem with interesting geometrical property, and span across ergodic theory, semiclassical analysis, spectral theory, and sub-Riemannian and contact geometry. All topics that can be

2 Bernoulli Institute Annual Report investigated from a geometric perspective by exploiting Hamiltonian systems and geodesic flows to describe their dynamical and spectral properties. Dr. Jacolien van Rij was appointed as tenure-track assistant professor and Rosalind Franklin Fellow in the Cognitive Modeling group in January 2018. The topic of her research is the acquisition and processing of (seemingly) ambiguous sentences, such as idioms. She is using a combination of computational simulations, experimental studies, and statistical methods to answer her research questions. After a postdoc position in Germany, she came to Groningen in 2016 to work on her NWO Veni project at the Faculty of Arts.

Prof.dr. J.B.T.M. Roerdink Scientific Director Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence

3 Bernoulli Institute Annual Report

4 Bernoulli Institute Annual Report

Contents

Introduction 1

Table of Contents 6

Governing body and support staff 7

List of scientific programmes and tenured scientific staff 9

Research schools and clusters 12

1 Algebra 15

2 Computational and Numerical Mathematics 25

3 Dynamical Systems, Geometry & Mathematical Physics 37

4 Probability and Statistics 59

5 Systems, Control and Applied Analysis 69

6 Distributed Systems 85

7 Fundamental Computing 101

8 Information Systems 107

9 Intelligent Systems 117

10 Scientific Visualization and Computer Graphics 147

11 Software Engineering 161

5 Bernoulli Institute Annual Report

12 Autonomous Perceptive Systems 177

13 Cognitive Modeling 191

14 Multi-Agent Systems 203

Colloquium Computer Science 2018 – List of Speakers 217

Colloquium Mathematics 2018 – List of Speakers 219

Colloquium Artificial Intelligence 2018 – List of Speakers 223

6 Bernoulli Institute Annual Report

Governing body and support staff

The information below reflects the situation after the establishment of the Bernoulli Institute on the 1st of June 2018, when the former Johann Bernoulli Institute for Mathematics & Computer Science (JBI) and the Institute for Artificial Intelligence & Cognitive Engineering (ALICE) became a single institute.

Scientific director Prof.dr. J.T.B.M. Roerdink

Scientific Board Prof.dr. N.A. Taatgen (chair) (professor of artificial intelligence, RUG) Prof.dr. B. Verheij (professor of artificial intelligence, RUG) Prof.dr. P. Avgeriou (professor of computer science, RUG) Prof.dr. J. Top (professor of mathematics, RUG) Prof.dr. M.K. Camlibel (professor of mathematics, RUG)

Management team fte J. de Jong-Schlukebir (policy officer) 0.6 M. Sanders (financial controller) 1.0 S. Costache (scientific coordinator) 0.5

International Advisory Panel (IAP) until June 1st Prof.dr. J. van Mill (chair) Prof.dr. M. Lenzerini Prof.dr. F. Vaandrager Prof.dr. J.A. Bergstra Prof.dr. W.Th.F. den Hollander

7 Bernoulli Institute Annual Report

International Advisory Panel (IAP) from June 1st Prof.dr. J.A. Bergstra (chair) Prof.dr. M. Lenzerini Prof.dr. W.Th.F. den Hollander Prof.dr. B. Nuseibeh Prof.dr. U. Schmid Prof.dr. W. van der Hoek Prof.dr. C. Bachoc

Administrative staff Secretaries of Research Institute D.J. Hansen (until 1-09-2018) 0.8 K.M.E. Schelhaas 0.8 E. Sietsema 0.8 H.M. Steenhuis 0.5 H.J. de Waard (until 30-09-2018) 1.0 S. van Wouwe 0.8

Address:

Postal address: P.O. Box 407 9700 AK Groningen

Visiting address: Bernoulliborg Nijenborgh 9 9747 AG Groningen The Netherlands

Tel : 050-3636533 Email : bernoulli.offi[email protected] Web : https://www.rug.nl/research/bernoulli

8 Bernoulli Institute Annual Report

List of scientific programmes and tenured scientific staff

Mathematics page

Programme 1 : Algebra 15 Prof.dr. J. Top Dr. A.V. Kiselev Dr. J.S. Muller¨

Programme 2 : Computational & Numerical Mathematics 25 Prof.dr.ir. R.W.C.P. Verstappen Dr.ir. F.W. Wubs Dr. C. Bertoglio Prof.dr. A.E.P. Veldman (em.)

Programme 3 : Dynamical Systems, Geometry & Mathematical Physics 37 Prof.dr. G. Vegter Prof.dr. E. Verbitskiy Prof.dr. H. Waalkens Dr. K. Efstathiou Dr. A. Sterk Dr. M. Seri Dr. D. Valesin Prof.dr. H.W. Broer (em.) Prof.dr.ir. H.S.V. de Snoo (em.) Prof.dr. A.C.D. van Enter (em)

Programme 4 : Probability & Statistics 59 Dr. Tobias Muller¨ Prof.dr. E.C. Wit Dr. M. Grzegorczyk Dr. W.P. Krijnen

Programme 5 : Systems, Control & Applied Analysis 69 Prof.dr. A.J. van der Schaft Prof.dr. K. Camlibel Prof.dr. H.L. Trentelman Dr. B. Besselink Prof.dr. S. Trenn Dr. A. Waters

9 Bernoulli Institute Annual Report

Computer Science

Programme 6 : Distributed Systems 85 Prof.dr. A. Lazovik Prof.dr. M. Aiello

Programme 7 : Fundamental Computing Science 101 Prof.dr. G.R. Renardel de Lavalette Dr. J.A. Perez´ Parra Prof.dr. W.H. Hesselink (em.)

Programme 8 : Information Systems 107 Prof.dr. D. Karastoyanova Dr. G. Azzopardi

Programme 9 : Intelligent Systems 117 Prof.dr.sc.techn. N. Petkov Prof. M. Biehl Dr. K. Bunte Dr. M.H.F. Wilkinson

Programme 10 : Scientific Visualization & Computer Graphics 147 Prof.dr. J.B.T.M. Roerdink Prof.dr. A. Telea Dr. J. Kosinka

Programme 11 : Software Engineering 161 Prof.dr. P. Avgeriou Dr. A. Ampatzoglou Dr. V. Andrikopoulos Dr. R. Smedinga

10 Bernoulli Institute Annual Report

Artificial Intelligence

Programme 12 : Autonomous Perceptive Systems 177 Prof.dr L.R.B. Schomaker Prof.dr. R. Carloni Dr. M. Wiering Dr. A. Meijster Dr. S.M. van Netten

Programme 13 : CognitiveModeling 191 Prof.dr. N.A. Taatgen Dr. J. Borst Dr. F. Cnossen Dr. J. van Rij-Tange Dr. J. Spenader Dr. M.K. van Vugt

Programme 14 : Multi-Agent Systems 203 Prof.dr. L.C.Verbrugge Prof.dr. B. Verheij Prof.dr. D. Grossi

11 Bernoulli Institute Annual Report

Research schools and clusters

Researchers of the BI participate in the following research schools and in all four NWO clusters:

1. Discrete, Interactive and Algorithmic Mathematics, Algebra and Number Theory (DIAMANT) Coordinating institution: CWI & TU/e & UL Director: Prof.dr. B. de Smit Participating BI programme(s): DSGMP and Algebra

2. Dutch Institute of Systems and Control (DISC) Coordinating institution: Delft University of Technology Director: Prof.dr H. Nijmeijer Participating BI programme(s): SCAA

3. Geometry and Quantum Theory (GQT) Coordinating institution: University of Utrecht Director: Prof.dr. C. Faber (UU) Participating BI programme(s): DSGMP

4. Nonlinear Dynamics of Natural Systems (NDNS+) Coordinating institution: University of Twente Chair: Prof.dr S.A. van Gils Participating BI programme(s): DSGMP

5. Stochastics - Theoretical and Applied Research (STAR) Coordinating institution: Technical University Eindhoven Eurandom Chair: Prof.dr E.A. Verbitskiy (UL) Participating BI programme(s): Probability and Statistics

6. The J.M. Burgers Centre for Fluid Dynamics Coordinating institution: Delft University of Technology Director: Prof.dr.G.J.F. van Heijst Participating BI programme(s): CNM

7. Institute for Programming Research and Algorithmics (IPA) Coordinating institution: University of Eindhoven Director: Prof.dr. W.J. Fokkink Participating BI programme(s): SE

12 Bernoulli Institute Annual Report

8. Advanced School of Computing and Imaging (ASCI) Coordinating institution: Delft University of Technology Director: Prof.dr.ir. H. Bal Participating BI programme(s): IS, SVCG

9. School of Behavioral and Cognitive Neurosciences (BCN) Coordinating institution: University of Groningen Director: Prof.dr. R.A. Schoevers Participating BI programme(s): SVCG and AI

13 Bernoulli Institute Annual Report

14 Bernoulli Institute Annual Report

1 Algebra

Group leader: Prof.dr. J. Top

Tenured staff (BI members) source fte Prof.dr. J. Top RuG 1.0

Tenure track source fte Dr. J.S. Muller¨ RuG 1.0 Dr. A.V. Kiselev RuG 1.0

Emeritus source fte Prof.dr. M. van der Put RuG 0.0

Postdocs source fte Dr. M. Derickx (from April till August) RuG 1.0 Dr. P. Kilic¸er (from September first) RuG 1.0

PhD students M.-P. Noordman RuG 1.0 (supervisor: Top) E. Ruiz Duarte (till September) Conacyt 1.0 (supervisor: Top) S. Gajovic RuG 1.0 (supervisor: J.S. Muller)¨ E. Kaya RuG 1.0 (supervisor: J.S. Muller)¨

Guests A. Cruz-Morales, National University of Colombia, Bogota, Colombia P. Beelen, Technical University of Denmark, Copenhagen, Denmark J-C. Lario, Polytechnic University of Catalonia, Barcelona, Spain M. Stoll, University of Bayreuth, Germany L. Furst,¨ University of Bayreuth, Germany C. Neurohr, University of Oldenburg, Germany S. Vemulapalli, University of California, Berkeley, USA C. Sanabria, University de los Andes, Bogota, Colombia P. Kilic¸er, University of Oldenburg, Germany

15 Bernoulli Institute Annual Report

16 Bernoulli Institute Annual Report

1.1 Research Program

Number theory and Algebraic geometry Arithmetic properties of elliptic curves over a number field or a function field, like the rank and generators of the Mordell-Weil group, are the subject of study. Also work is done on applications to Diophantine equations, coding theory and arithmetic algebraic geometry; in particular a study of the number of rational points on curves over finite fields, and determining the rational points on curves over a number field. Moreover, the history and the algebraic geometry of various series of geometrical models is studied. Geometry of differential equations This concerns algebraic, analytic (e.g., multisummability) and algorithmic aspects of linear differential and linear difference equations; differential Galois theory and its applications, in particular to symbolic (algorithmic) solvability of equations; (Lie) symmetries of non-linear differential equations; isomonodromy and in particular the six Painleve´ equations; nonlinear first order equations, algebraic theory in positive characteristic. Moreover, developing and applying algebraic, geometric, and algorithmic techniques to nonlinear partial differential equations of Mathematical Physics. This research centers at the Kontsevich deformations of Poisson structures and deformation quantization of Poisson field models, as well as at a noncommutative extension of the Batalin–Vilkovisky approach to quantization in models of fundamental interactions.

1.2 Overview of scientific results

J. Top With M. van der Put and M.P. Noordman, a project on algebraic solutions of first order differ- ential equations via methods from algebraic geometry (in particular, generalized Jacobians) is expected to result in a paper soon. A project on variational methods applied to Painleve´ differential equation, jointly with Van der Put and P. Acosta-Humanez´ (University Simon´ Bol´ıvar, Barranquilla, Colombia), resulted in a manuscript which is still unpublished. Top and Van der Put continued their collaboration with C. Sanabria Malagon (Bogota), concerning the construction of linear differential operators with a given finite matrix group as Galois group. A paper resulting from this is being submitted. The project (based on work with former bachelor’s students) on number theoretical aspects of the classical Poncelet Closure Theorem, resulted in a publication in the International Journal of Number Theory. A text written jointly with his former PhD student A.S.I. Anema (a section of it is in fact based on the bachelor’s project of Anne Tuijp) dealing with properties of the Hesse pencil, was published in SIGMA.

17 Bernoulli Institute Annual Report

The text written jointly with I. Polo-Blanco (University of Cantabria, Santander, Spain), on history and use of geometric models illustrating the classification of cubic curves, was accepted for publication in the Mathematische Semesterberichte. Top started a collaboration with S. Tafazolean (Campinas, Brazil) resulting in a manuscript describing certain hyperelliptic maximal curves. This is expected to be published in 2019. J.S. Muller¨ The joint project with J. Balakrishnanan, N. Dogra, J. Tuitman and J. Vonk on the rational points of the split Cartan modular curve at level 13 was completed and the resulting manuscript was submitted to the Annals of Mathematics. Various projects which may be regarded as spin-off or generalizations or variations of the project above, are currently leading to lots of collaborations. Mathematicians working with Muller¨ on aspects of this, include J. Balakrishnan, A. Besser, F. Bianchi, N. Dogra, K. Kedlaya and J. Vonk. Muller¨ (as promotor and daily supervisor) at the University of Oldenburg finished the PhD project of Christian Neurohr, who successfully defended his thesis in March. A project with Raymond van Bommel and David Holmes (Leiden) resulted in a general algorithm for computing canonical heights on Jacobians using regular models. The joint manuscript describing this was submitted to a journal. Together with his Master student C. Stumpe, Muller¨ worked on bounds for the difference between the naive and the canonical height on elliptic curves. A paper resulting from this was accepted by Acta Arithmeticae. Finally, an ongoing joint project with A. Besser on the computation of local height pairings without models was continued. A. Kiselev Arthemy Kiselev and Ricardo Buring provided a detailed explanation why the graph orientation morphism takes cocycles in the Kontsevich graph complex to Poisson cocycles, that is, to infinitesimal symmetries of Poisson brackets on arbitrary finite-dimensional affine manifolds. The outline is based on a text by C. Jost (Stockholm), itself follwing a paper by T. Willwacher (ETH Zurich)¨ in Invent. Math., which in turn refers to a sketch in the seminal paper by Kontsevich (1996). Kiselev read an IMPRS lecture course on this subject to PhD students and postdocs at the Max Planck Institute for Mathematics (Bonn, Germany) in December. Also, Kiselev phrased the construction of the orientation morphism in purely combinatorial terms of (un)oriented graphs. This was the topic of Arthemy’s invited talk at the workshop ‘Homotopy algebras, deformation theory and quantization’ (September 2018, Be¸dlewo, Poland). In Be¸- dlewo, R. Buring presented a joint poster about high order expansion ? mod o¯(~4) of the

18 Bernoulli Institute Annual Report

Kontsevich star-product; Buring spoke about the construction of orientation morphism at the 32nd International colloquium GROUP32 on group-theoretical methods in Physics (July 2018, CVUT Prague, Czech Republic). In December, Buring pronounced a joint talk about the graph cocycles at the international workshop ‘Symmetries and integrability of equations of Mathematical Physics’ in the Institute of Mathematics NAS (Kiev, Ukraine). Kiselev was invited to give a talk there but his visit to Kiev was impossible because of the Ukraine entry ban. Together with N. J. Rutten, Kiselev wrote the proof of many technical lemmas which in their totality guarantee that the Kontsevich unoriented graph complex with the vertex-expanding differential is well defined modulo the ‘zero’ graphs of any size. This substantiation had been lacking in the literature. Nina Rutten reported the proof at the 27th Winter school on Mathematical Physics (January 2018, Janske´ Lazn´ e,ˇ Krknose,ˇ Czech Republic) and at the colloquium GROUP32 in July. When the long paper by Kiselev about a field-theoretic extension of the Kontsevich formal noncommutative symplectic supergeometry was published in J. Geom. Phys. on 5 April, Kiselev resumed the study of quantum Riemannian geometries realised on dynamical tilings of smooth or affine manifolds. He presented this problem at the colloquium GROUP32 in Prague. In May and October, Kiselev visited the Johannes Gutenberg Universitat¨ (Mainz, Germany) where R. Buring works on a PhD project with Kiselev; both trips were supported by the Mathematical Institute in Mainz. Arthemy continued his collaboration with Andrey Krutov from the Independent University of Moscow (IUM). In July and December, Arthemy visited the IUM, Ivanovo State Power university (ISPU), and Lebedev IP RAS in Moscow for scientific collaboration. To everyone’s extreme sorrow, Mr. A. B. Sokolov – who had been working at ISPU on a PhD project under supervision by Kiselev – passed away on January 18. M. Derickx Maarten Derickx worked in our group for only 4 months, during a short gap between his time in Leiden and his appointment at MIT. In Groningen he continues work with S.J. Edixhoven (Leiden) and M.P. Noordman, expected to result in a joint paper by the three of them. He also presented various lectures and supervised two bachelor’s projects. Jointly with B. Mazur and S. Kamienny he wrote and published a paper on 17-torsion of elliptic curves, which appeared in the journal Contemporary Mathematics. P. Kılıc¸er Initiated by a collaboration in a Women in Numbers meeting, Pinar Kıc¸ıcer and others wrote a paper on modular invariants for genus three hyperelliptic curves. This was submitted to Research in Number Theory.

19 Bernoulli Institute Annual Report

Moreover, a joint paper with Labrande, Lercier, Ritzenthaler, Sijsling, and Streng was pub- lished in Acta Aritmeticae. In November she visited D. Casazza (Madrid) and they started a collaboration on explicit constructions of certain class fields. Eduardo Ruiz Duarte Duarte finished his PhD project on theory and applications of genus two curves. He defended his thesis in May. Apart from the thesis itself, two manuscripts are being prepared based on it: jointly with Top, a text explaining the part of the thesis which extends Manin’s elementary proof of Hasse-Weil to the case of genus two. This was submitted to the Indian Journal of Mathematics. Jointly with Noordman, a second text is being prepared, in which a primality test is proposed based on arithmetic with certain genus two curves. This is expected to be finished next year. Marc-Paul Noordman Noordman, Boix (Beersheva) and Top wrote a manuscript extending the notion of ‘level’ appearing in the theory of differential operators in characteristic p > 0. This is expected to be submitted in 2019. His work with Duarte and with Van der Put and Top was already mentioned above. He gave various lectures, including one at an international Rational Points conference on Schiermonnikoog. He also participated in a summer school on p-adic differential equations in Poland. Stevan Gajovic Gajovic continued his PhD project with Muller.¨ He gave various talks in a local seminar. Enis Kaya Kaya continued his work on p-adic heights and special instances of a p-adic Birch & Swinnerton- Dyer conjecture. He gave various talks, and participated in some international conferences (among others the Building Bridges meeting, and a meeting on modular forms in Luxemburg).

1.3 Research subjects

E. R. Duarte: Genus 2 curves, applications to cryptography and number theory. M.P. Noordman: algebra and geometry of differential equations, including in positive charac- teristic. S. Gajovic: Integral points on curves. E. Kaya: Arithmetic of abelian varieties, in particular algorithms for p-adic heights. A. Kiselev: Geometry of differential equations, geometry of fundamental interactions, inte- grable systems, BV- and deformation quantisation, brackets, mathematical physics.

20 Bernoulli Institute Annual Report

M. Derickx: arithmetic of modular curves, torsion of elliptic curves. P. Kılıc¸er: CM theory, arithmetic of genus two and three curves. J.S. Muller¨ : Computational arithmetic geometry, in particular algorithms for rational points on curves and their Jacobians and for arithmetic intersection theory. J. Top: arithmetical algebraic geometry, in particular: elliptic curves and surfaces, curves over finite fields and over function fields, history of geometrical models, Galois representations, number theory; differential equations.

1.4 Publications

Articles in peer-reviewed journals

A.S.I. Anema, J. Top, and A. Tuijp, Hesse pencils and 3-torsion structures, SIGMA • (Special Issue on Modular Forms and String Theory in honor of Noriko Yui), 14 (2018), 102, 13 pages.

Buring R., Kiselev A. V., Rutten N. J. (2018) Poisson brackets symmetry from the • pentagon-wheel cocycle in the graph complex, Physics of Particles and Nuclei 49:5 (2018), Supersymmetry and Quantum Symmetries’2017, 924–928.

P. Kılıc¸er, H. Labrande, R. Lercier, C. Ritzenthaler, J. Sijsling, and M. Streng, Plane • quartics over Q with complex multiplication. Acta Arith., 185 (2018), no. 2, 127–156.

Kiselev A. V. (2018) The calculus of multivectors on noncommutative jet spaces, J. Geom. • Phys. 130 (2018), 130–167.

Kiselev A. V., Krutov A. O. On the (non)removability of spectral parameters in Z2-graded • zero-curvature representations and its applications, Acta Appl. Math., 42 p. (in press) doi:10.1007/s10440-018-0198-6.

J. Los, T. Mepschen, and J. Top, Rational Poncelet, Internat. J. Number Theory, 14 • (2018), 2641–2655.

J.S. Muller¨ , Applying the Mordell-Weil sieve: Appendix to Quadratic Chabauty and • rational points, I: p -adic heights, by Jennifer S. Balakrishnan and Netan Dogra, Duke Math. J., 167 (2018), 1981–2038.

21 Bernoulli Institute Annual Report

Chapters in books

Kiselev A. V., Krutov A. O., Wolf T. (2018) Computing symmetries and recursion oper- • ators of evolutionary super-systems using the SSTOOLS environment, in: Nonlinear Systems and Their Remarkable Mathematical Structures 1 (N. Euler, ed.) CRC Press, Boca Raton FL, USA, 390–407.

Articles in peer-reviewed proceedings

Buring R., Kiselev A. V., Rutten N. J. Infinitesimal deformations of Poisson bi-vectors • using the Kontsevich graph calculus, J. Phys.: Conf. Ser. 965 (2018), Proc. XXV Int. conf. ‘Integrable Systems & Quantum Symmetries’ (6–10 June 2017, CVUT Prague, Czech Republic), Paper 012010, 1–12.

Other publications

Griffioen S. F., Kiselev A. V. (2016) Painting new lines: maximizing color difference in • metro maps, published in 2016 in The Mathematical Intelligencer; reprinted 31 January 2018 by Scientific American.

1.5 External funding and collaboration

External funding

The PhD position of Eduardo Ru´ız Duarte (until August 2018) is funded by Conacyt (Mexico). The PhD position of S. Gajovic is funded for 3 years by a DFG grant awarded to Muller.¨ The PhD position of E. Kaya is funded for 2 years by an NWO grant awarded to Muller.¨ Together with M. Stoll, Muller¨ is jointly supervising the PhD project of L. Furst,¨ funded for 3 years by DFG. Furst¨ is currently employed at the University Bayreuth and will transfer to Groningen in 2019. Kiselev won a visitor grant from the IHES´ (Bures-sur-Yvette, France), which will allow for his collaboration with M. Kontsevich in November–December 2019. External collaboration See 1.2

22 Bernoulli Institute Annual Report

1.6 Further information

Top served on the PhD evaluation committee for M. Mornev (Leiden, supervisors Bas Edix- hoven and Lenny Taelman, defence Feb. 13th). Top is a board member of the Foundation Compositio Mathematicae, and an editor of the Newsletter of the European Mathematical Society and of the journal Indagationes Mathematicae. In 2017 Top served on the NWO EW committee evaluating TOP1 grants and TOP2 grants. Top and Muller¨ and their PhD students and postdoc participate in the North German Alge- braic Geometry Seminars (NoGAGS, a collaboration between Gottingen,¨ Hannover, Berlin, Hamburg, Bremen and Groningen), and they are involved in the NWO-cluster DIAMANT. Top gave an invited lecture in the BIRS-CMO conference “Rational and Integral Points via Analytic and Geometric Methods” in Oaxaca, Mexico and he participated in the ANTS conference in Madison, USA. Top continues to organize lectures popularizing math, e.g., for the Groningen “college car- rousel”. Moreover (as in previous years) he organizes the regional math olympiad at the RUG, followed by several training sessions for gifted high school children. Muller¨ gave an invited talk at a workshop on the Arithmetic of hyperelliptic curves at the ICTP in Trieste (Italy). He also spoke in the Seminar Algebra, Geometry and Number theory at the University of Leiden. Together with M. Bright, D. Schindler and A. Smeets he organized an international conference on the island of Schiermonnikoog, which took place in July 2018. Kiselev served on the PhD promotion committee for Remko Klein (VSI Groningen, supervisor D. Roest). Kiselev gave three seminar talks: at the Algebra, Geometry and Physics seminar (MPIM Bonn, Germany) in November, as well as the Quantum Field Theory seminar in the I. E. Tamm Theoretical Department of the Lebedev IP RAS (Moscow, Russia) and seminar on Lie algebras, Riemannian geometry and Mathematical Physics at the Independent University of Moscow in December. Involved in the NWO-cluster Geometry & Quantum Theory (GQT), Arthemy attended two conferences and graduate schools which were organised by that cluster in June and November. Through the entire year 2018, Kiselev was the co-organizer of the (Johann) Bernoulli Institute mathematics colloquium. Kiselev is a member of the Faculty Library committee at RuG. Kiselev is an editorial board member in J. Nonlinear Mathematical Physics; he also reviewed many publications for Math. Rev.

23 Bernoulli Institute Annual Report

A recreational-math paper by S. Griffioen and A. Kiselev (outlining a strategy to choose new colour(s) to paint the new line(s) on a metro map, originally published by The Mathematical Intelligencer 38(1) in 2016) was reprinted on 31 January 2018 by Scientific American.

24 Bernoulli Institute Annual Report

2 Computational and Numerical Mathematics

Group leader: Prof.dr.ir. R.W.C.P. Verstappen

Tenured staff (BI members) source fte Prof.dr.ir. R.W.C.P. Verstappen RUG 1.0 Dr.ir. F.W. Wubs RUG 1.0

Tenure track source fte Dr. C. Bertoglio RuG 1.0

Emeritus Prof.dr. A.E.P. Veldman

Postdocs Dr. ir. P. van der Plas STW 1.0 Dr. P. Cifani RUG 1.0

PhD students S. Baars MSc NWO 1.0 (supervisor: Wubs) H. Carrillo MSc Conicyt/RUG 1.0 (supervisors: Bertoglio, Waters) J. Garay MSc RUG 1.0 (supervisor: Bertoglio) D. Nolte MSc Conicyt 1.0 (supervisor: Bertoglio) J. Parekh MSc NWO 1.0 (supervisor: Verstappen) R.A. Remmerswaal MSc STW 1.0 (supervisor: Veldman, Verstappen) Ir. J.H. Seubers STW 1.0 (supervisors: Van der Plas, Veldman) D.-L. Sun MSc NSFC/RUG 1.0 (supervisor: Carpentieri, Verstappen) L.B. Streher MSc RUG 1.0 (supervisor: Verstappen)

25 Bernoulli Institute Annual Report

eScience Research Engineer J. Hidding MSc eScience Center 0.5

Guests E. Mulder (Utrecht, 12 months) M.H. Silvis (12 months) N. Valle Marchante (UPC Barcelona, 3 months) H. Saghi (Iran, 3 months) F.X. Trias (UPC Barcelona, 1 month)

26 Bernoulli Institute Annual Report

2.1 Research Program

With the continuing progress in numerical mathematics and computer technology, the impact of computer simulation on society is rapidly increasing. Our group specializes in numerical algorithms for the simulation of fluid dynamics and transport phenomena (Computational Fluid Dynamics CFD). On the one hand research is focussed on basic advancement of numerical algorithms; on the other hand - through extensive cooperation with external research groups - these methods are made available to advance knowledge in other (applied) areas of science and technology. Turbulence In most applications, the Navier-Stokes equations do not provide a trackable model for turbulent flow. Therefore, finding a coarsed-grained description is one of the main challenges to turbulence research. A most promising methodology for that is large- eddy simulation (LES). The basic idea of LES is that the large scales of motion remain virtually unchanged, whereas the calculation of all small-scale turbulence for which numerical resolution is not available is avoided. This keeps the computational effort within reasonable limits, but a price is paid in terms of accuracy. To improve the accuracy, we perform research into scale-truncation models for large-eddy simulation. The mathematical rationale behind our approach focusses on approximations that preserve the underlying PDE structure as well as on regularizations that truncate the nonlinear interactions with small scales of motions. Free-surface flow and fluid-structure interaction The free-surface flow research concerns application in maritime and coastal engineering. Numerical simulation methods are developed to predict hydrodynamic wave loading on offshore platforms and coastal constructions. The basic tool is the in-house developed simulation method ComFLOW. The ComMotion project focusses on the dynamic interaction with (floating) moving and (elastically) deforming objects (fluid-structure interaction). The related SLING project focusses on two-phase liquid sloshing in LNG tanks. Fluid-structure interaction is also central in our research concerning bio-medical fluid dynamics (with UMCG). Inverse problems in blood flows Here we develop improved means for quantifying flows with emphasis on the cardiovascular system. Improvements are made through: (a) developing models incorporating the neglected physical domain in an computationally efficient and numerically robust fashion, and (b) to mathematically formulate and numerically solve inverse problems in order to parametrize these models using data coming medical images, which also cope the drawbacks intrinsic to the specific imaging modalities (e.g. aliasing, undersampling). Sparse-matrix solvers The repeated solution of large systems of equations in most sim- ulation methods makes the quest for improved matrix solvers another major research area. In-house a number of multilevel preconditoners have been developed. For general systems, we designed MRILU (Matrix renumbering ILU) and VBARMS (Variable Block Algebraic

27 Bernoulli Institute Annual Report

Recursive Multilevel Solver). MRILU can be applied to discretizations of coupled PDEs. It particularly performs well for convection-diffusion equations. VBARMS almost automatically exploits any available block structure during the factorization, achieving increased throughput during the computation and improved reliability on realistic applications. Additionally, the special purpose multilevel preconditioner HYMLS (Hybrid Multilevel solver) is designed to meet the incompressiblity constraint efficiently. Both VBARMS and HYMLS are parallelized using MPI. Typical application areas of the solvers are fluid flow, structural problems, electro magnetics and bifurcation analysis. Bifurcation analysis Here the emphasis is on numerical methods for investigating the bifurcation behaviour of fluid flow, also in the presence of noise. Applications range from academic to real world problems like the lid-driven cavity problem and the global ocean circulation, respectively.

2.2 Overview of scientific results

Large eddy simulations (LES) of turbulence resort to coarse-grained models of the small scales of motion for which numerical resolution is not available. LES has advanced so far that it could be used as a design tool in the not too distant future. However, these simulations are still rife with errors and uncertainties; particularly, the model that describes the unresolved- resolved interactions is a major source of uncertainty. Therefore we have started to analyze the nonlinear propagation of uncertainties coming from the turbulence model in large eddy simulations. Here the main focus is on eddy-vsicosity models. Further, our joint work with Stanford University (Center for Turbulence Research) and UPC Barcelona focused on the performance of minimum-dissipation models for rotating channel flows and definition of length scales for large eddy simulations on anisotropic computational grids. In the ComMotion project, extensions of the ComFLOW simulation method are being designed featuring moving and deforming objects. For the numerical coupling between the solid mechanics and the fluid dynamics a quasi-simultaneous method has been developed that is stable for any added-mass ratio. The dispersive absorbing boundary conditions now allow for current. Applications are e.g. free-fall life boats, floating buoys and elastically deforming objects. Much effort has been spent in testing the code and preparing it for final release to the project partners. The RUG contribution to the SLING project focusses on the simulation of flow instabilities, where capillary forces and turbulence play a role. Their simulation requires high numerical accuracy. By reducing the spurious velocities, the calculation of the curvature- related surface tension effects has been improved. Along the free surface, the solution is allowed to be discontinuous in the tangential velocity component and, correspondingly, in the pressure gradient. The illustration clearly shows this discontinuous character of the velocity

28 Bernoulli Institute Annual Report

field.

Figure 1: Two-phase simulation of a breaking wave. The color shows the velocity, which changes rather abruptly across the free surface.

For numerical inverse problems in blood flows, we have formulated boundary conditions that, parametrized with imaging data, allow to cope for uncertainties in the computational domain typically coming from the limited resolution of the images. We also have developed reduced models to represent distal vasculature in a simpler way and even more with less assumptions. We also have accomplished flow velocity reconstruction methods in magnetic resonance imaging more robust to aliasing than the state of the art. In the research on sparse-matrix solvers for ocean models, the parallel tailored THCM solver has been extended with ice and results are underway. A long-standing problem in our solver for CFD problems, HYMLS, has been resolved. Performance results are underway. For numerical bifurcation analysis the developed solver for generalized Lyapunov equations has been applied to study stochastic marine ice sheet variability. A paper has been finalized and accepted for publication. The Lyapunov solver was meant to study transition probabilities. However, this appeared not to be possible. Therefore focus has shifted to trying to apply the Adaptive Multilevel Splitting (AMS) method and the Genealogical Particle Analysis (GPA) method to geostrophic flows. Until now, these methods have only be applied to small problems. The parallel implementation of the dynamical orthogonal field method (DO method) for general systems of PDEs has been completed and is currently tested on a quasi-geostrophic (QG) model. In this method, higher moments of probability distributions can be computed than what is possible with the Lyapunov solver. A paper on pattern formation in a 2D and 3D Turing problem has been written and has been accepted for publication.

29 Bernoulli Institute Annual Report

2.3 Research subjects

S. Baars: efficient solution of generalized Lyapunov equations. C.A. Bertoglio: blood flows modelling, in particular boundary conditions, and related inverse problems from medical imaging, in particular Magnetic Resonance Imaging. H.J. Bandringa: simulation of complex flows in maritime applications. Y. Bu: matrix factorization methods for Markov chains, preconditioners for iterative solution of linear systems of equations. H. Carrillo: blood flows inverse analysis from velocity MRI. P. Cifani: numerical simulations of multi-phase flows. H. Carrillo: blood flows inverse analysis from velocity MRI. X.M. Gu: Krylov subspace methods for solving non-Hermitian linear systems with application to fractional differential equations. S. Kotnala: efficient solution of stochastic PDEs. D. Nolte: blood flows inverse analysis from velocity MRI. P. van der Plas: local grid refinement for free-surface flow simulation. R.A. Remmerswaal: two-phase liquid sloshing in LNG cargo tanks. J.H. Seubers: interaction between extreme waves and floating bodies. Z. Li Shen: matrix solvers for Markov chains problems. M.H. Silvis: models for the larger eddies in turbulent flow. W. Song: numerical linear algebra for bifurcation analysis on high-performance computers. L.B. Streher: applications of large-eddy simulations of turbulence. D.-L. Sun: numerical linear algebra methods in computational nanophotonics A.E.P. Veldman: modeling and simulation of fluid flows in engineering applications, free- surface flows, fluid-structure interaction. R.W.C.P. Verstappen: mathematics of Computational Fluid Dynamics (CFD), modeling and simulation of turbulence. F.W. Wubs: preconditioners for sparse systems and Lyapunov equations in CFD; application to stability and bifurcation analysis and the study of the influence of noise on bifurcation points.

30 Bernoulli Institute Annual Report

2.4 Publications

Dissertations

Y. Bu. A class of linear solvers based on multilevel and supernodal factorization. • Promotor: prof.dr. A.E.P. Veldman; Co-promotor: dr. B. Carpentieri. University of Groningen, 3 July 2018.

Articles in scientific journals

C. Bertoglio, A. Caiazzo, Y. Bazilevs, M. Braack, M. Esmaily, V.L. Gravemeier, A. Mars- den, O.E. Pironneau, I.E. Vignon-Clementel and W.A. Wall, Benchmark problems for numerical treatment of back fow at open boundaries. International Journal for Numeri- cal Methods in Biomedical Engineering 34, (2018). C. Bertoglio, R. Nunez,˜ F. Galarce, D. Nordsletten, and A. Osses. Relative pressure estimation from 4D velocity measurements in blood flows: state-of-the-art and new approaches. International Journal for Numerical Methods in Biomedical Engineering, 34 (2018). Y. Bu, B. Carpentieri, Z. Shen and T. Huang, Multilevel inverse-based factorization preconditioner for solving sparse linear systems in electromagnetics. Applied Computa- tional Electromagnetics Society Journal 33, 160–163 (2018). D. Castellana, H.A. Dijkstra, and F.W. Wubs, A statistical significance test for sea-level variability Dynamics and Statistics of the Climate System. 3 (2018). P. Cifani, J.G.M. Kuerten and B.J. Geurts, Highly scalable DNS solver for turbulent bubble-laden channel flow. Computers & Fluids 172, 67–83 (2018). X.M. Gu, T.Z. Huang, G. Yin, B. Carpentieri, C. Wen and L. Du, Restarted Hessenberg method for solving shifted nonsymmetric linear systems. Journal of Computational and Applied Mathematics 331 166–177 (2018). J. Hormann,¨ C. Bertoglio, M. Kronbichler, M. Pfaller, R. Chabiniok, and W. Wall. An adaptive Hybridizable Discontinuous Galerkin approach for cardiac electrophysiology. Int. J. Num. Meth. Biomed. Eng., 34 (2018). M. Li, X.M. Gu, C. Huang, M. Fei and G. Zhang, A fast linearized conservative fi nite element method for the strongly coupled nonlinear fractional Schr’odinger´ equations. Journal of Computational Physics 358 256–282 (2018).

31 Bernoulli Institute Annual Report

T.E. Mulder, S. Baars, F.W. Wubs & H. Dijkstra, Stochastic marine ice sheet variability, Journal of Fluid Mechanics 843 748-777 (2018).

W. Rozema, R.W.C.P. Verstappen, J.C. Kok and A.E.P. Veldman. Low-dissipation simulation methods and models for turbulent subsonic flow. Archives of Computational Methods in Engineering 1–32 (2018).

Z.L. Shen, T.Z. Huang, B. Carpentieri, C. Wen and X.M. Gu, Block-accelerated aggre- gation multigrid for Markov chains with application to Pagerank problems. Communi- cations in Nonlinear Science and Numerical Simulation 59 472–487 (2018).

W. Song, F. Wubs, J. Thies, S. Baars, Numerical bifurcation analysis of a 3D turing- type reaction–diffusion model Communications in Nonlinear Science and Numerical Simulation 60, 145-164 (2018).

D.L. Sun, B. Carpentieri, T.Z. Huang and Y.F. Jin, A spectrally preconditioned and initially deflated variant of the restarted block GMRES method for solving multiple right-hand sides linear systems. International Journal of Mechanical Sciences 144, 775–787 (2018).

D.L. Sun, T.Z. Huang, Y.F. Jing and B. Carpentieri, A block GMRES method with deflated restarting for solving linear systems with multiple shifts and multiple right-hand sides. Numerical Linear Algebra with Applications 25 e2148 (2018).

R. Verstappen, On closing large-eddy simulations. ERCOFTAC Bulletin 116 26–31 (2018).

R. Verstappen, How much eddy dissipation is needed to counterbalance the nonlin- ear production of small, unresolved scales in a large-eddy simulation of turbulence? Computers & Fluids 176 276–284 (2018).

Articles in conference proceedings

M.R. Pfaller, M.C. Varona, J. Lang, C. Bertoglio, W.A. Wall, Parametric model order reduction and its application to inverse analysis of large nonlinear coupled cardiac problems, arXiv:1810.12033, 2018.

M.H. Silvis, R. Verstappen, Nonlinear subgrid-scale models for large-eddy simulation of rotating turbulent flow. In Direct and Large Eddy simulation XI, M. Salvetti al (eds) pp 129-134 Springer, 2018.

32 Bernoulli Institute Annual Report

M.H. Silvis, R. Verstappen, Constructing physically consistent subgrid-scale models for large-eddy simulation of incompressible turbulent flows, In Turbulence and Interactions TI2015, M.O. Deville et al (eds) 241-248, Springer, 2018.. L.B. Streher, M.H. Silvis, R. Verstappen, Mixed modeling for large-eddy simulation: the minimum-dissipation-Bardina model, arXiv:1806.11317, 2018 Also: Proceedings 7th European Conference on Computational Fluid Dynamics, R. Owen et al (eds) page 335-345, 2018. L.B. Streher, P. Cifani, R.W.C.P. Verstappen, Application of the minimum-dissipation model to turbulent bubble-laden flows, In: Proceedings 12th Int Ercoftac Symposium on Engineering Turbulence Modeling and Measurement ETMM12, 6 pp. Montpelier, 26-28 Sept. 2018. D.L. Sun, B. Carpentieri, T.Z. Huang, Y.F. Jing and S. Naveed, Variants of the block- GMRES method for solving linear systems with multiple right-hand sides. 2018 Inter- national Workshop on Computing, Electromagnetics, and Machine Intelligence (CEMi), 21-24 Nov., 2018, Stellenbosch, South Africa, IEEE Xplore, 2018. P. van der Plas, A.E.P. Veldman, J. Helder, K.-W. Lam. Adaptive grid refinement for two-phase offshore applications. 37th Int. Conf. Ocean, Offshore and Arctic Eng., Madrid, 17-22 June, 2018. paper OMAE2018-77309, 10 pages. A.E.P. Veldman, H. Seubers. M. Hosseini Zahraei, P. van der Plas and P.R. Wellens. Strong quasi-simultaneous coupling for fluid-structure interaction in offshore applica- tions. NUTTS, 30 Sep. - 2 Oct., 2018, Cortona (I), 6 pages. A.E.P. Veldman, H. Seubers, S. Matin Hosseini Zahraei, P. van der Plas, P.R. Wellens. R.H.M. Huijsmans. Preventing the added-mass instability in fluid-solid coupling for offshore applications. 37th Int. Conf. Ocean, Offshore and Arctic Eng., Madrid, 17-22 June, 2018. paper OMAE2018-77308, 10 pages. R. Verstappen, A discrete scale-truncation model for LES. In Direct and Large Eddy simulation X, D. Grigoriadis et al (eds) pp 157-164 Springer, 2018.

2.5 External funding and collaboration

Most of our PhD projects are being funded externally from national and international resources. We summarize the situation:

The PhD students David Nolte and Hugo Carrillo are mainly funded by the Chilean • Research Council (Conicyt).

33 Bernoulli Institute Annual Report

The ComMotion project concerning the interaction between extreme waves and moving • or deforming objects, in cooperation with TU Delft and MARIN, is funded by STW, the Topsector Water and several offshore companies. Its budget is around 1 MEuro. Researchers involved are Veldman, Van der Plas and Seubers.

Our contribution to the STW Perspectief SLING project on liquid sloshing in LNG tanks • started in 2016 (Remmerswaal). It is a cooperation with the Dutch technical universities and a worldwide industrial consortium, and has a total budget of around 6 MEuro.

Most of the turbulence research is funded by the Free Competition of NWO EW • (Silvis), the Netherlands Enterpise Agency (RVO), Topsector Water (Bandringa). The exchange with UPC Barcelona is mainly paid by the spanish Ministerio de Economia y Competitividad and the Generalitat de Cataluyna.

The Ubbo Emmius Fund and the German Aerospace Laboratory each sponsor two years • of a PhD project on numerical algorithms for exascale computers (Song). The NWO program Mathematics of Planet Earth supports a PhD-project on the study of efficient Lyapunov solvers for the oceanographic applications (Baars).

The Ubbo Emmius Fund and the National Natural Science Foundation of China sponsor • a PhD project on the design and parallelization of multilevel matrix solvers for partial differential equations (Bu). The Ubbo Emmius Fund and the University of Electronic Science and Technology of China sponsor two years each of two PhD projects on the design and parallelization of multilevel matrix solvers for solving partial differential equations (Gu and Shen). They also fund a follow-up project on numerical linear algebra methods and applications in computational nanophotonics (Sun).

The Accelerating Scientific Discovery programme of the Netherlands eScience Center • funds our new project on stochastic multiscale climate models (Baars and Mulder)

NWO’s Joint CSER and eScience programme for Energy Research funds the project • parallel-in-time methods for the propagation of uncertainties in wind-farm simulations (Parekh)

Societal relevance As indicated above, part of our PhD and MSc research is carried out in physical or technological applications. Close cooperation exists wit several university research laboratories, with all Dutch Technological Institutes (GTI’s), and with several industries: multi-nationals as well as small and medium enterprises.

34 Bernoulli Institute Annual Report

(Inter)national collaboration Various bilateral contacts exist with research groups inside and outside the Netherlands, leading to e.g. joint PhD projects, participation in summer schools, traineeships for Master’s students and/or to joint publications.

The research on free-surface flows and hydrodynamic wave loading, focussed around • the ComFLOW development, is embedded in world-wide joint-industry projects with main partners the Maritime Research Institute MARIN, TU Delft, Deltares, FORCE Technology Norway and several offshore companies and shipyards. The related SLING project on two-phase liquid slohing in LNG tanks is a cooperation with the three Dutch technical universities and supported by a large industrial consortium.

The research on methods for bifurcation analysis for ocean circulation models is carried • out in close cooperation with prof. H.A. Dijkstra from the Institute for Marine and Atmospheric Research (IMAU) in Utrecht. The project on numerical linear algebra for bifurcation analysis on high-performace computers is a cooperation with Dr. J. Thies and Dr. A. Basermann from DLR (Cologne). Furthermore there is cooperation with Prof. M. Bollhoeffer from the Technical University of Braunschweig on the solution of sparse linear systems by multilevel ILU preconditioners. With Dr. Hochstenbach from TUE and Prof. V. Simoncini (University of Bologna) we cooperate on the numerical solution of Lyapunov equations.

The research in Magnetic Resonance Imaging is performed with experimental partners • in Santiago de Chile, Berlin and Amsterdam. We have ongoing collaborations in the field of cardiovascular modeling with the Technical University of Munich and INRIA France. In Groningen, we are doing retrospective clinical studies at UMCG to study relation between the blood vessel properties and clinical outcomes in pulmonary hypertension patients.

Our turbulence research comprises cooperation with the Universitat Politecnica` de • Catalunya in Barcelona, TU Munchen,¨ TU Berlin, KU Leuven and Stanford University. We also work together with MARIN and NLR.

(Inter)national PhD committees

Veldman participated in 2 PhD defense committees at TU Delft. Verstappen participated in 4 PhD defense committee (UPC Barcelona, TU Delft and 2xRUG) Wubs participated in 2 PhD defense committe (RUG).

35 Bernoulli Institute Annual Report

2.6 Further information

Bertoglio was invited speaker at the following scientific events: 8th Int. Conf. & Summer School on Multiscale Modeling (Santiago, Chile); Modeling & Optimization of the Cardiovas- cular System (Magdeburg, Germany); III Science without Frontiers (Groningen, Netherlands); Mathematical Modeling in Hemodynamics (Saint-Etienne, France).

Veldman is a scientific consultant of the National Aerospace Laboratory NLR (Amsterdam) and he is a member of the Advisory Board of the Maritime Research Institute MARIN (Wageningen). Also, he has a part-time emeritus position at the University of Twente. He is chairman of the jury for the KIVI Charles Hoogendoorn Award. Further, he is on the editorial board of Journal of Engineering Mathematics and Journal of Algorithms and Computational Technology. He delivered lectures at the ECCOMAS CFD Conference 2018 in Glasgow (Scotland), the 37th International Symposium on Offshore Mechanics and Arctic Engineering (Madrid, Spain) and the Numerical Towing Tank Symposium (Cortona, Italy).

Verstappen is member of the NWO committee for Scientific Use of Supercomputers (WGS), the Steering Committee of the Special Interest Group Large Eddy Simulation of ERCOFTAC (European Research Community On Flow Turbulence And Combustion), the Steering Com- mittee of the Groningen Engineering Center (GEC), the Board of Projectleaders of the JM Burgers Center and chair of the JM Burgers Contact Group Computational Fluid Dynamics. Further, he is deputy director of the Bachelor’s and Master’s programmes in (Applied) Mathe- matics at RUG and board member of the Undergraduate School for Science and Engineering (USSE). He gave keynote lectures at the 10th International Conference on Compuational Fluid Dynamics ICCFD10 (Barcelona, Spain) and the 5th International Conference on Turbulence and Interactions (Martinique). Together with F.X. Trias he organized a minisymposium at the 7th ECCOMAS Conference on Computational Fluid Dynamics (Glasgow, UK). He is on the editorial board of Computers and Fluids.

Wubs is member of the Woudschoten committee organizing the annual meeting of Dutch- Flemish numerical analists. Wubs and Baars visited DLR Cologne several times to collaborate with Thies. Several PhD students gave presentions at conferences and workshops. Also talks and posters at the “J.M. Burgersdag” (Lunteren) and the Dutch-Flemish Conference on Scientific Computing (Woudschoten) can be mentioned.

36 Bernoulli Institute Annual Report

3 Dynamical Systems, Geometry & Mathematical Physics

Group leader: Prof.dr. G. Vegter

Tenured staff (BI members) source fte Prof.dr. G. Vegter RuG 1.0 Prof.dr. E. Verbitskiy RuG/Leiden 0.2 Prof.dr. H. Waalkens RuG 1.0

Tenure track source fte Dr. K. Efstathiou RuG 1.0 Dr. M. Seri RuG 1.0 Dr. A. Sterk RuG 1.0 Dr. D. Valesin RuG 1.0

Emeriti source fte Prof.dr. H.W. Broer RuG 0.0 Prof.dr. A.C.D. van Enter RuG 0.0 Prof.dr.ir. H.S.V. de Snoo RuG 0.0

PhD students Gabriel Leite Baptista NWO 1.0 (supervisors: D. Valesin and A. van Enter) P. L. Barrios Pantoja Ubbo Emmius and 1.0 (supervisors: D. Valesin, M. Jara and A. van Enter) IMPA Matthijs Ebbens UG PhD Scholarship Program 1.0 supervisors: G. Vegter and A. Sterk J. Gao CSC scholarship 1.0 (supervisors: H.W. Broer and K. Efstathiou) J. Hidding (Kapteyn Institute/RuG) NWO 0.5 (supervisors: R. Van de Weijgaert and G. Vegter) D. L. van Kekem RuG 1.0 (supervisor: Sterk)

37 Bernoulli Institute Annual Report

Bohuan Lin CSC scholarship 1.0 (supervisors: K. Efstathiou and H. Waalkens) N. Martynchuk (until September2018) RUG 1.0 (supervisors: H.W. Broer and K. Efstathiou) Eric Pap (since September 2018) UG PhD Scholarship Program 1.0 supervisors: D. Boer (VSI) and H. Waalkens G. Wilding (since February 2018) DSSC COFUND 1.0 (supervisors: R. Van de Weijgaert (promoter), G. Vegter (promoter) and K. Efsathiou (co- promoter)) Y. Zhang CSC scholarship 1.0 (supervisors: H.W. Broer and K. Efstathiou)

Guests Caio Alves, Universitat¨ Leipzig, Germany Jussi Berhndt, TU Graz, Oostenrijk Rodrigo Bissacot, University of Sao Paulo, Brazil Alexey Bolsinov, Loughborough, UK Conrado Costa, Leiden University Holger Dullin, Univ. Sydney, Australia Heinz Hanssmann,¨ University of Utrecht, the Netherlands Seppo Hassi, University of Vaasa, Finland Sonja Hohloch, Univ. Antwerp, Belgium Frank den Hollander, University of Leiden, the Netherlands Iordan Iordanov, INRIA Nancy, France Christof Jung, UNAM Cuernavaca, Mexico Andreas Knauf, FAU Erlangen, Germany Christof Kuelske, Ruhr-Universitaet Bochum, Germany Max Lein, Advanced Institute of Materials Research, Tohoku University, Sendai, Japan Bernardo de Lima, UFMG, Brazil Jacek Miekisz, University of Warsaw, Poland Thomas Mountford, EPFL, Switzerland Arnaud Le Ny, Laboratoire d’Analyse et de Mathematiques Appliquees, France Balazs Rath, BME, Hungary Frank Redig, University of Delft, Netherlands Rodrigo Ribeiro, PUC Chile, Chile

38 Bernoulli Institute Annual Report

Wioletta M. Ruszel, University of Delft, Netherlands Renato Soares dos Santos, NYU Shanghai, China Monique Teillaud, INRIA Nancy, France Ferdinand Verhulst, University of Utrecht, the Netherlands Mats Vermeeren, Institut fur¨ Mathematik, Technische Universitat¨ Berlin, Germany San Vu Ngoc, Univ. Rennes I, France Florian Wagener, University of Amsterdam, the Netherlands

39 Bernoulli Institute Annual Report

40 Bernoulli Institute Annual Report

3.1 Research program

The research of the group Dynamical Systems, Geometry and Mathematical Physics covers a broad and diverse spectrum of subjects in the fields of fundamental, applied and computational dynamical systems theory, classical, statistical and quantum mechanics and their interfaces in the light of dynamics, and theoretical and applied aspects of geometry with many connections to dynamical systems theory. The interest of dynamical systems theory is the behaviour of systems that evolve in time. This first of all concerns the long-term behaviour which comprise stationary, periodic, multi-periodic and chaotic dynamics, but also transient behaviour is of interest. Moreover bifurcations or transitions between asymptotic states – in particular transitions between regular and chaotic motions – under variation of parameters are of great importance. We develop mathematical tools using methods from analysis, geometry and measure theory to grasp, study and develop the structures involved. Moreover, we develop methods to detect and understand the dynamics in specific models, employing numerical and graphical tools and computer algebra. Many applications are from the field of mechanics. This concerns the motion of point masses like planets and their satellites in celestial mechanics, and also the motion of atoms and molecules which again can be described as point masses or rigid bodies. Here also relativistic or quantum effects may play a role. This is a wide area with great outreach, also in the direction of life sciences. If the number of constituent particles is huge then such systems are best described by statistical means. Statistical mechanics deals with the question of how global observables, like temperature, can be explained from the microscopic behaviour. There is a close relationship with dynamical systems theory in particular with regard to random and chaotic behaviour and the so-called non-equilibrium systems. Mathematical physics is the encompassing discipline of all the above and still larger areas of theoretical physics. The research is very geometrical in nature. It involves many tools from differential and computational geometry where the research in the field of geometry has many facets also in its own right.

Dynamical Systems Theory The discipline of Dynamical Systems is concerned with mathematical models for deterministic time evolutions. A simple example is derived from the oscillator, which generally only displays periodic dynamics. If subject to periodic driving or to coupling with another oscillator, it can illustrate many parts of the Dynamical Systems research program.

One possible state of the system is resonance, where the combined system assumes one globally periodic state, the frequency of which is an integer combination of the individual periodic motions. Another possible state is multi- or quasi-periodicity, where the individual periodic motions combine in a rationally independent way. When coupling three oscillators, a third possible combined state exists, where a continuous range of frequencies is present: this

41 Bernoulli Institute Annual Report is the state of chaos.

The occurrence of resonance, quasi-periodicity and chaos as well as the transitions or bifurca- tions in between, is the central theme of research in the current Dynamical Systems program – not only for a few coupled oscillators but for a wide class of systems.

The questions posed vary from fundamental to applied, where the focus can be on different classes of systems. Examples of this are the world of general ‘dissipative’ systems with a finite-dimensional state space, the classes of Hamiltonian or reversible systems or systems with a very low-dimensional state space. Also concrete examples are being studied, where sometimes numerical or symbolic algorithms have to be developed. The mathematics of these different levels strongly interact. For instance, in order to know what to look for in a special case, one has to know beforehand what can be expected and what is logically possible.

There is cooperation with groups in other sciences on the analysis of specific systems. This concerns the Department of Engineering, University of Bristol (A. Champneys), the University of Auckland (B. Krauskopf, H.M. Osinga), the University of Eindhoven (H. Nijmeijer), the Department of Applied Mathematics and Analysis of the University of Barcelona (C. Simo´ and A.` Jorba) as well as the University of Utrecht (O. Diekmann, H.A. Dijkstra, H. Hanßmann, T. Opsteegh, F. Verhulst), Boston University (R.L. Devaney, T. Kaper), the University of Exeter (M.P. Holland and D.B. Stephenson) and the Radboud University Nijmegen (M. Krupa). Various PhD students and postdocs at all these institutions are involved as well. The theoretical work is also internationally oriented and involves intensive cooperation with the universities of Dijon (R. Roussarie), Ohio State University (M. Golubitsky), the Universite´ de Marseille (J.C. Poggiale and S. Troubetskoy), the Russian Academy of Sciences (M.B. Sevryuk) and other places, resulting in joint publications on a regular basis.

Geometry and Geometric Approximation The research topics in geometry fall within several subfields of Geometry, like Differential Geometry and Singularity Theory on the more theoretical side, and Computational Geometry on the more applied side. In collaboration with other local and international groups in Mathematics and Computer Science, this research is also focused on applications in Geometric Modeling, Dynamical Systems, and Astrophysics. The goal is to obtain constructive and certified methods for the study of geometric and topological structures arising in a wide range of scientific problems. The approximation of shapes in various representations by ‘simple’ geometric objects, like polyhedral objects or piecewise quadratic surfaces, is one of the central topics of our research program. This field has attracted a lot of interest from researchers in Computer Aided Geometric Design, where mainly numerical aspects of the problem are emphasized. However, our scope is rather different in that we focus on topological correctness, a criterion often disregarded in applications: the topology of the approximating object should

42 Bernoulli Institute Annual Report be the same as that of the original shape. We also apply techniques from Algebraic and Computational Topology to Data Analysis in Cosmology. This project involves an investigation of the Cosmic Web, the space-filling foamlike pattern permeating the Universe. The Cosmic Web represents the most striking example of a complex geometric pattern in nature and defines the cosmic environment in which galaxies emerge. In both the observed galaxy distribution as well as in computer simulations of cosmic structure formation we see matter accumulate in walls, filaments and dense compact clusters, surrounding vast near empty voids. The research is carried out in collaborations with the Cosmology group of R. van de Weygaert at the Kapteyn Astronomical Institute of the University of Groningen, with the INRIA-groups DataShape (Boissonnat, Dyer, Wintraecken) at Sophia Antipolis and GAMBLE (Lazard, Teillaud) at Nancy, France, with C.K. Yap of the Courant Institute of Mathematical Sciences at New York University, and with A. Chattopadhyay at IIIT-Bangalore, India.

Statistical Mechanics and Stochastics Statistical mechanics forms an integral part of the research in the group. Both equilibrium and non-equilibrium questions are considered. In comparison with dynamical systems, the emphasis is on systems with infinitely many, rather than finitely many, degrees of freedom. Links between the two can be fruitfully studied in the thermodynamic limit, where one can consider the asymptotics of systems with a large (but finite) number of degrees of freedom. Such questions often lead to the use of stochastic methods.

The general aim of the stochastics part of the program is to understand interacting stochastic systems on a mathematical level. Even when the interaction is local, such systems typically exhibit a complex global behavior, with a spatial long-range dependence resulting in phase transitions. In this picture phase transitions are characterized by discontinuous behavior of the possible states of the system as a function of external parameters. The equilibrium properties of such systems can often be described by Gibbs measures. These are probability measures on the set of possible states of the system given in terms of the energy resulting from local interactions. For specifically tuned values of the parameters, there can be more than one Gibbs measure, and Gibbs measures can be highly non-trivial. In one direction of research we are focusing on these. Moreover we are also interested in the time-evolution of such measures. It was discovered a few years ago that time-evolved measures may lose (and recover) their Gibbsian nature as a function of time. Related to this is the preservation of the Gibbs property under discretisations and spatial coarse-graining. We are trying to approach this phenomenon in case studies. In a related line of research we are investigating continuous interfaces. We also study the behaviour of large systems in inhomogeneous external fields. We have also studied one-dimensiona wetting phenomena, and the connection between Gibbs

43 Bernoulli Institute Annual Report measures and g-measures, that is between two-sided and one-sided continuity properties of conditional probabilities. This has led to a study of various aspects of low-dimensional long-range models.

Percolation theory is another field in which the study of phase transitions and other statistical- mechanical phenomena have been very fruitful. The object of study of percolation theory are the connectivity properties of random sets obtained after some stochastic procedure is applied to a lattice or graph. We are currently working on models of percolation in which this stochastic procedure presents some form of long-range dependence, a theme that has received considerable attention in recent years. We also investigate how stable long-range models models are in the sense of changing inifinite range to finite but sufficiently large range in some models in 3 and more dimensions.

We are also working on models for temperature-dependence of n vector models, on diffraction − theory with M. Baake (Bielefeld) and D. Lenz (Jena), on the theory of disordered systems and on models for metastability such as bootstrap percolation and in general on examples of phase transitions of physical and conceptual interest. We are also starting to develop the theory in relationship with the theory of networks.

Finally, we conduct research on the statistical mechanics of systems that are out of equilibrium. Metastability is an ubiquitous phenomenon involving systems whose transition to equilibrium is hindered by the presence of energy barriers or “traps”. We are conducting research on the metastable behavior of a model called the contact process, which can both be taken as a toy model for out-of-equilibrium dynamics and as a model for the evolution of an infection in a biological population. We consider this model on different classes of random networks and study how the topological properties of the network affect the transition to equilibrium.

The work is done in various collaborations, including in the last years M. Baake (Bielefeld), R. Bissacot (Sao Paulo), L. Coquille (Grenoble), H. Duminil-Copin (Geneve` and IHES, Bures- sur-Yvette), R. Fernandez´ (Utrecht/Shanghai), A. Fey (Delft) F. Redig (Nijmegen-Delft), A. Le Ny (Paris), W.Th.F. den Hollander (Leiden), C. Kulske¨ (Bochum), A.A. Opoku (Sunyani, Ghana), R. van der Hofstad (Eindhoven), H.G. Dehling (Bochum), S. Shlosman (Marseille), V. Zagrebnov (Marseille), S. Romano (Pavia), K. Netocny´ (Prague), S. Roelly (Potsdam), D. Lenz (Jena), T. Hulshof (Eindhoven), J-C. Mourrat (ENS Lyon), T. Mountford (EPFL Lausanne), Balazs´ Rath´ (BME Budapest), B.N.B. de Lima (Belo Horizonte, E.Saada(CNRS and Paris- Descartes), M. Mourragui (Rouen), Anna Levit (Vancouver), Bruno Schapira (Marseille), B.Kimura, W.M. Ruszel (Delft), C.Spitoni, (Utrecht), Leonardo T. Rolla (Buenos Aires and Shanghai).

Mathematical Physics

44 Bernoulli Institute Annual Report

Several directions of the themes mentioned above originate from or are oriented towards research questions in physics and astronomy. This concerns both fundamental research questions and also applications. The orientation towards physics is also reflected by several collaborations with the Van Swinderen Institute for Particle Physics and Gravity, the Kapteyn Astronomical Institute and the Zernike Institute for Advanced Materials at the University of Groningen and with physics institutions outside of Groningen. Besides the directions already mentioned under themes above the following research subjects belong to the mathematical physics activities in the group.

Integrable Systems.– In the theory of Hamiltonian dynamics the geometry of the phase space plays an important role, in particular the bundle structure of invariant tori in integrable systems. The nontriviality of such bundles is studied by methods from differential geometry and algebraic topology, where this has led to the development of monodromy and Chern classes. These results have a counterpart in semi-classical quantum theoretical approximations where quantum monodromy helps to explain certain spectral defects. Molecular and atomic systems are also studied from this point of view. This research is carried out by H. W. Broer, H. Waalkens, and K. Efstathiou in cooperation with various groups including the Universite´ du Littoral, Dunkerque (B. Zhilinski´ı, D. Sadovski´ı), University of Calgary (R.H. Cushman), Utrecht University (H. Hanßmann), University of Catania (A. Giacobbe), Universite´ de Bourgogne, Dijon (D. Sugny, P. Mardesiˇ c),´ Sydney University (H.R. Dullin) and University of Erlangen (A. Knauf).

Optics and Semiclassics.– In the limit of short wavelengths wave theories like optics and quantum mechanics reduce to ray optics and classical mechanics, respectively. This limit is cumbersome and involves several open mathematical questions on the one hand and on the other hand also provides besides deep conceptual insights also powerful computational tools. One example is the monodromy of torus bundles in integrable systems already mentioned above. Another example is given by optical micro resonators. New fabrication techniques allow one to build lasers and optical resonators on a microscopic scales. Here the light is trapped in micro-scale cavities utilizing the principle of total internal reflection. Such micro resonators have great potential for miniaturing spectroscopic devices and diagnostic tools. For many applications, it is of crucial importance to optimize the quality factors of these devices together with the directionality of the optical output. This can be achieved by, e.g., a suitable choice of the morphology of the cavity boundary. Significant insight into the output directionality for a given cavity shape can be obtained on the level of the ray dynamics from studying the corresponding billiard map. Combining this with techniques from semiclassical quantum mechanics (short wavelength asymptotic) leads to the design of cavities with laser modes which are both long lived and directional. Further ideas like perturbing circular disk cavities by a point scatterer are also exploited. This research is carried out by H. Waalkens in collaboration with C. Dettmann, M. Sieber (both Bristol University) and G. Morozov

45 Bernoulli Institute Annual Report

(University of the West of Scotland).

Classical and Quantum Transport.– Many questions in dynamical systems theory can be phrased as transport problems. This in particular concerns the study of reaction type dynamics which is associated with transport through phase space bottlenecks. In the simplest case such bottlenecks are induced by saddle type equilibrium points and this is the case best understood up to now. Generally speaking phase space bottlenecks are realized normally hyperbolic invariant manifold and the transport through them is mediated by their stable and unstable manifolds. The research has a wide range of applications ranging from chemical reactions to ballistic electron transport problems and to capture and escape problems in celestial mechanics. Using methods from semiclassical analysis like the Weyl symbol calculus the theory can be carried over to quantum transport problems. The present research in the fields addresses questions like the construction of normally hyperbolic invariant manifolds in more general settings, bifurcations of normally hyperbolic invariant manifolds and the quantum mechanical manifestations of these. The research is carried out in collaboration with U.¨ C¸ iftc¸i (Namık Kemal University), H. R. Dullin (Sydney University), R. Schubert and S. Wiggins (both Bristol University), G. Ezra (Cornell University) and C. Jung (UNAM Cuernavaca). Furthermore, problems of transport of electrons in metals perturbed by slowly varying magnetic fields can be understood in a semiclassical sense by means of adiabatic techniques coupled with magnetic pseudo-differential calculus. This involves both the study of classical dynamics of integrable systems, and the analysis of the integrated density of states of Schrodinger¨ operators with long range perturbations of different kinds. A new line of work has recently been established in this direction by M. Seri in collaboration with L. Parnovski and A. Sobolev (University College London, UK), D. Elton (Lancaster University, UK), M. Lein (Institute of Advanced Materials, Sendai, Japan) and G. De Nittis (Pontifical Catholic University of Chile).

Contact Hamiltonian Mechanics.– Contact mechanics is the odd-dimensional counterpart of symplectic mechanics. Historically it has been mostly consider for the study of classical thermodynamics from a geometric point of view, but is acquiring more and more prominence in recent years as the natural setting for quasi-Hamiltonian systems, like the ones describing the law of motion of bodies with damping. Research carried in this area involves the derivation of new numerical methods that can preserve the contact structures, and the analytical analysis of the phase space structure. Applications to electrical circuits, cosmology and quantum mechanics are also considered. This research is carried out mainly by M. Seri in collaboration with A. Bravetti (CIMAT, Guanaguato, Mexico) and M. Vermeeren (TU Berlin, Germany). A PhD student is expected to join the collaboration in 2019.

Dynamics on networks.– Many applications can be modelled as networks of coupled oscillators. Research carried in the group is concerning the study of the dynamics of such networks. Of

46 Bernoulli Institute Annual Report particular interest are synchronization phenomena. Models of couple oscillators networks are, e.g., used to describe the circadian rhythms of organisms. The research is carried out mainly by K. Efstathiou in collaboration with I. Hoveijn who has close contact to the Chronobiology group in Groningen.

Extreme Events.– Using a combination of tools from dynamical systems theory and statistics the occurrence of extreme events in dynamical systems like the weather system is studied. This research is carried out by A.E. Sterk in a collaborations with M.P. Holland and D.B. Stephenson (both at University of Exeter). This collaboration has grown of out of the EU network Complexity-NET which concluded in 2013.

Operator Theory.– The main focus here is the extension theory of symmetric and sectorial operators in Hilbert spaces and in spaces with an indefinite metric. This extension theory is closely connected to mathematical physics (explicitly solvable models, singular perturbations), to system theory (the realization in terms of transfer functions) and to analysis (moment problems, interpolation problems, differential operators, canonical systems). Furthermore, Lebesgue type decompositions and Radon-Nikodym derivatives are being studied in the context of pairs of bounded linear operators.

The research is done in collaboration with the following group of mathematicians: J. Behrndt (Graz), T. Berger (Hamburg), S. Hassi (Vaasa), J.Ph. Labrousse (Nice), A. Sandovici (Iasi), F.H. Szafraniec (Krakow), C. Trunk (Ilmenau), H. Winkler (Ilmenau).

3.2 Overview of scientific results

Dynamical Systems Theory In KAM theory several projects are running in cooperation with Broer. A monograph entitled ‘Quasi-periodic bifurcation theory: the geometry of KAM’, co-authored by Hanßmann and F.O.O. Wagener (UvA) is in preparation.

Geometry Vegter, together with his former PhD student Wintraecken and postdoc Dyer, continued the work on non-degeneracy criteria for Riemannian simplices based on comparison with simplices in spaces of constant sectional curvature. The previous bound in the condition depended on the absolute curvature of the space. In our recent work we show that the degeneracy condition depends on the variation of the curvature, that is, if the curvature lies within two bounds, then

47 Bernoulli Institute Annual Report the condition depends on the difference of the bounds. This work has led to two papers, which have been accepted for publication in 2019. Together with Teillaud and Iordanov (INRIA), Ebbens and Vegter extended their earlier work on Delaunay triangulations of point sets on hyperbolic surfaces. We determined explicit expressions for the systoles, i.e., the lengths of shortest geodesics, of symmetric Riemann surfaces of arbitrary genus. Using this result we determined the space complexity of the optimal algorithm for Delaunay triangulations of a large class of surfaces. This work is an extension of our earlier work. It has been presented at the workshop Curves and Surfaces, Arcachon, France, in 2018. In collaboration with Parlier (U. Luxembourg) we proved that Delaunay triangulations of a large class of hyperbolic Riemann surfaces of genus g have complexity at least linear in g. This is in striking contrast with the complexity of arbitrary triangulations of such surfaces, which may be of order √g. G. Vegter and K. Efstathiou together with their PhD student G. Wilding, and K. Nevenzeel and R. van de Weygaert (Kapteyn) have worked on the topological data analysis of ΛCDM cosmological models.

Statistical Mechanics About the Gibbsian-non-Gibbsian program, around which the previous PhD projects of Iacobelli, Ermolaev, Opoku and Ruszel were centered there were a number of developments. In this program one studies which measures can and cannot be written as a Gibbs measure for an effective Hamiltonian. This is often done in physics, although it turns out to be not always mathematically justified. Various examples of physical interest occur, e.g., in Renormalization Group theory, in the theory of disordered systems and in the study of non-equilibrium problems, and in the theory of discrete approximations.

In 2017 Verbitskiy continued the study of renormalized Gibbs states and algebraic properties of dynamical systems. The paper on ergodic properties of random continued fractions has appeared in Nonlinearity). Jointly with K. Schmidt (Vienna) and D. Lind (Seattle), he continued investigating renormalization of algebraic dynamical systems. He completed a joined project with R. Fernandez (Utrecht) and S. Berghout (Leiden) on the relation between g and Gibbs measures, the paper has been accepted for publication in Ergodic Theory & Dynamical Systems.

With H. Duminil-Copin (Geneve` and IHES, Bures-sur-Yvette ) and T. Hulshof (Eindhoven) van Enter finished his study on the existence of correction terms to sharp thresholds in anisotropic bootstrap percolation models. A joint paper appeared in Prob.Th.Rel. Fields, and a short review of the result in a conference proceedings.

48 Bernoulli Institute Annual Report

With joint PhD student E. Endo, cosupervised with R. Bissacot (Sao Paulo), two papers appeared, one together with Bissacot and A. Le Ny (Paris) on the difference between g- measures versus Gibbs measures and the role of entropic repulsion in Comm. Math. Phys., another one with Bissacot, B. Kimura and W.M.Ruszel (Delft) in Ann.H. Poicare´ on the persistence of phase transitions of one-dimensional long-range Ising models in inhomogeneous fields. Van Enter publshed a paper in J. Stat. Phys. with L.Coquille (Grenoble), Le Ny and Ruszel on the possible existence of interface Gibbs measures in two dimensions. Moreover he wrote a paper on metastability in one-dimensional long-range models with Kimura, Ruszel and C. Spitoni (Utrecht), which has been accepted for publication. He submitted a short review on the work with Bissacot,Endo and Le Ny for a conference proceedings, which was accepted in the mean time, and he started a project with J. Miekisz(Warsaw) and H. Koisuvalo (Vienna) on the construction of one-dimensional balanced quasicrystalline ground states for pair interactions.

In percolation, in collaboration with Bernardo N. B. de Lima (Universidade Federal de Minas Gerais) and Leonardo T. Rolla (Universidade de Buenos Aires and NYU Shanghai), D. Valesin has studied the critical curve for a multi-range percolation model on oriented trees. The results obtained shed some light into monotonicity issues in percolation. An article has been accepted for publication in Random Structures and Algorithms. This line of research has been further developed together with Bernardo de Lima and the PhD student Reka´ Szabo;´ finer results concerning the shape of the critical curve, as well as the asymptotic behavior of the supercritical percolation cluster, have been obtained. An article is in preparation. Again in percolation, with Balazs´ Rath´ (BME Budapedt), D. Valesin has proved that the upper stationary distribution of the supercritical contact process on the d dimensional lattice (with d 2) exhibits a phase transition with respect to the infection rate, provided that the ≥ range of interaction in the process is taken large enough. An article is in preparation. In metastability, D. Valesin has collaborated with Bruno Schapira (Aix-Marseille Universite)´ to prove the existence of a limiting exponential rate in the function that describes the duration of activity of the supercritical contact process on finite graphs. The existence of such a constant had previously been obtained only for deterministic graphs, such as lattice boxes; in the work at hand, it has been established for a broad class of random graphs. A paper has been submitted.

Algebraic Dynamical Systems Verbitskiy has continued the joint project with T. Shirai (Fukuoka, Japan), on the relation be- tween spanning trees of infinite graphs and equal entropy algebraic systems. In a joint project with K. Schmidt (Vienna) and D. Lind (Seattle), on renormalization of algebraic dynamical systems, first results has been obtained. In particular, it has been shown that the renormal-

49 Bernoulli Institute Annual Report ized principal actions remain principal. Existence of scaling limits of the corresponding polynomials is under investigation.

Integrable Systems K. Efstathiou and H. Waalkens in collaboration with N. Martynchuk and H. Dullin (Sydney) considered the problem of scattering monodromy in the two-center problem. In collaboration with H. Dullin they considered the problem of monodromy for the isotropic 3-DOF harmonic oscillator. Two papers have been submitted. With H. Dullin H. Waalkens has studied mon- odromy in the Kepler problem. A paper was published in Physical Review Letters with the distinction of an Editors’ suggestion.

K. Efstathiou continued his work on integrable Hamiltonian systems. A paper with N. Mar- tynchuk and H. W. Broer on Chern classes and monodromy has been submitted. With H. Hanßmann (Utrecht) and A. Marchesiello (Prague) they have submitted a paper on the bifurca- tions and geometry of 3-DOF oscillator systems with 1:1:-2 resonance.

Classical and Quantum Transport Waalkens has published a paper on roaming reaction dynamics together with V. Krajnˇak´ in the Journal of Mathematical Chemistry. A paper in collaboration with G. Ezra (Cornell) and S. Wiggins (Bristol) concerning the application to the HCN isomerization problem is in preparation. With C. Jung (UNAM Cuernavaca) H. Waalkens studies the break up of normally hyperbolic invariant manifolds in the application of ionization of hydrogen in a circularly polarised electric field. A paper is in preparation. Max Lein visited M. Seri in September, to kickstart some of the work, and prepare to organise a Workshop in relation to Semiclassical Theory of Metallic Transport. A longer follow-up visit is planned for February 2019. M. Seri visited London for two weeks to continue working with L. Parnovski and D. Elton on the asymptotic expansion for the integrated density of states of Landau operators with quasi-peridic perturbations.

Contact Hamiltonian Mechanics M. Vermeeren (TU Berlin, Germany) visited the Bernoulli Insititute in December to finalise a paper jointly with M. Seri and A. Bravetti (CIMAT, Guanaguato, Mexico). M. Seri and A. Bravetti submitted an application for a NWO Visiting Researcher, to have A. Bravetti visit the institute for a month in 2019 and proceed with their research program.

Optics and semiclassics Waalkens continued his collaboration with C. Dettmann, M. Sieber (both Bristol University)

50 Bernoulli Institute Annual Report and G. Morozov (University of the West of Scotland) on optical micro cavities with point scatterers. A paper is in preparation.

Exceptional points. Waalkens has started a collaboration with D. Boer from Van Swinderen Institute on exceptional points which are degeneracies in parameter families of non-hermitian operators. Together they supervise the master student E. Pap. The collaboration concerns in particular the study of non-abelian phenomena due to coexisting multiple exceptional points. A paper suggesting a waveguide experiment to probe non-abelian phenomena has been published in Physical Review A. Various other papers mainly concerning the mathematical background are in preparation.

Extension theory Lebesgue type decompositions and Radon-Nikodym derivatives in the context of pairs of operators and forms are being studied with Hassi; abstract operator decompositions are worked on with Labrousse, Sandovici, and Winkler. Normal, not necessarily intermediate, extensions of symmetric operators are being studied with Hassi and Szafraniec. The work on maximal sectorial extensions with Hassi, Sandovici, and Winkler is being continued. The decomposition of linear relations in a finite-dimensional space is an ongoing project with Berger, Trunk, and Winkler. A monograph on boundary value problems is in preparation with Behrndt and Hassi.

Extreme events A paper on the predictability of extremes in dynamical systems by A.E. Sterk and his collabo- rator M.P. Holland (University of Exeter) was published in the journal Dynamics and Statistics of the Climate System by Oxford University Press. The results presented in this paper has lead to new questions on the predictability of extremes in intermittent systems; this topic will be pursued in future papers. Also the predictability of extremes in the context of imperfect models is a highly interesting direction for future research.

Dynamics on networks K. Efstathiou worked with J. Gao on the dynamics of second order oscillator networks. One paper has been published in Phys. Rev. E. A second paper has been submitted and a third one is in preparation. He also worked on networks with Winfree dynamics with Y. Zhang and I. Hoveijn; a paper is under preparation.

Casimir Oscillators H. Waalkens works in collaboration with G. Palasantzas from the Zernike Insitute for Ada- vanced Materials on the dynamics of Casimir oscillators. At separations below 100 nm, Casimir forces strongly influence the actuation dynamics of microelectromechanical systems (MEMS). For many applications, the roughness of the surface elements significantly influences

51 Bernoulli Institute Annual Report the qualitative dynamics of the oscillator. In particular chaotic motion can lead to malfunction of MEMS oscillators. The occurrence of chaos is studied using Melnikov functions for differ- ent materials and experimental setups. Papers have been published in Physical Review E and European Physics Journal B.

PhD research Martynchuk has defended his thesis on the geometry of integrable Hamiltonian systems in September 2018. In total, 4 papers have been accepted for publication, and one has been submitted. Wilding uses topological data analysis methods to study the geometric properties of the distribution of matter in the early universe. He is supported by a DSSC COFUND scholarship. A paper is in preparation with his supervisors R. van de Weygaert (Kapteyn), G. Vegter, and K. Efstathiou. Gao is supported by a scholarship from the China Scholarship Council. His research concerns the dynamics of networks of second order oscillators. A paper has been published, a second paper has been submitted, and a third paper is under preparation. Zhang studies network models for circadian rhythms. She is supported by a 2-year scholarship from the China Scholarship Council. A paper is currently in preparation with her supervisor K. Efstathiou and I. Hoveijn. Lin started his research on non-compact and fractional monodromy in September 2017 with his supervisors K. Efstathiou and H. Waalkens. He is supported by a scholarship from the China Scholarship Council. Dirk van Kekem finished his PhD research under the supervision of A.E. Sterk and successfully defended his thesis on 12 October 2018. The research has resulted in four joint papers which were published in leading journals on dy- namical systems. Swier Garst, a maths teacher at RGO Middelharnis, finished his PhD research under the supervision of J.M. Aarts (deceased), H.W. Broer, and A.E. Sterk and successfully defended his thesis on 19 October 2018. The thesis was based on two joint papers. The prelimi- nary stage of the research was financially supported by a “Leraar in Onderzoek” grant from the NWO. Eric Endo defended his thesis, supervised by A.C.D. van Enter, R. Bissacot (Sao Paulo) and D. Valesin on June 29. He obtained a cum laude degree. Two chapters were based on work with his supervisors Bissacot and van Enter, respectively with A Le Ny, and with B. Kimura and W.M.Ruszel, another chapter was based on a joint work with D. Valesin on random graphs.

3.3 Research subjects

H.W. Broer: perturbation and KAM-theory, bifurcation theory, non-integrable and resonance phenomena, applications of singularity theory, exploration of complicated systems. A.C.D. van Enter:lattice statistical mechanics and thermodynamic formalism, Gibbs-non- Gibbs transitions, bootstrap percolation, nonlinear vector models, disordered systems and spin-glasses, metastates and chaotic size-dependence, non-crystalline long-range order, inho- mogeneous fields. K. Efstathiou: integrable and near-integrable Hamiltonian systems, applications of Hamil- 52 Bernoulli Institute Annual Report tonian mechanics in physical systems, generalized monodromy, network dynamics with applications to biology and engineering. Y.M. Ebbens: Discrete geometry of hyperbolic surfaces. E.O. Endo: Ising systems in inhomogeneous fields, random graphs, g-measures. J. Gao: Dynamics on networks of second-order oscillators. J. Hidding (Kapteyn Institute/RuG): Mathematical simulation of cosmic structure forma- tion. D.L. van Kekem: bifurcation analysis of large-scale systems, such as the Lorenz-96 system. B. Lin: classical, non-compact, and quantum fractional monodromy. N. Martynchuk: geometry of integrable Hamiltonian systems. M. Seri: spectral theory, spectral geometry, contact mechanics, sub-Riemannian geometry. H.S.V. de Snoo: extension and realization theory with their applications to analytical prob- lems; Lebesgue type decompositions and Radon-Nikodym derivatives. A.E. Sterk: numerical exploration of dynamical systems, statistics and predictability of ex- treme events in dynamical systems, applications to climate models, applications to biological models, analysis in infinite-dimensional spaces; in particular related to (partial) differential equations. D. Valesin: interacting particle systems, percolation theory, metastability, random graphs. G. Vegter: certified geometric approximation, computational topology, discrete differential geometry. E. Verbitskiy: lattice statistical mechanics, thermodynamic formalism, Gibbs-non-Gibbs transitions, dynamical systems and time-series prediction. H. Waalkens: theoretical and application oriented aspects of Hamiltonian systems including integrable systems, monodromy, reaction type dynamics, invariant manifolds, normal forms, and semiclassical quantum mechanics (short wavelength asymptotics) with applications to micro lasers and quantum reaction dynamics. G. Wilding: Topological data analysis of cosmological simulations. Y. Zhang: Network models for circadian rhythms.

3.4 Publications

Dissertations

– E.O. Endo, Gibbs measures for models on lines and trees, Promoters: R. Bissacot and A.C.D. van Enter, co-promoter: D. Valesin, Bernoulli Institute, University of Groningen, 12 October 2018, 106 pages.

– N. Martynchuk, On monodromy in integrable Hamiltonian systems, Promotors: H.W. Broer,

53 Bernoulli Institute Annual Report

K. Efstathiou, Bernoulli Institute, University of Groningen, September 21 2018, 121 pages.

– D.L. van Kekem, Dynamics of the Lorenz-96 Model: Bifurcations, Symmetries and Waves, Promotor: H.W. Broer, co-promotor: A.E. Sterk, Bernoulli Institute, University of Groningen, 12 October 2018, 193 pages.

– S.H.P. Garst, Dynamics Amidst Folding and Twisting in 2-Dimensional Maps, Promotor: H.W. Broer, co-promotor: A.E. Sterk, Bernoulli Institute, University of Groningen, 19 October 2018, 96 pages.

Articles in scientific journals

– E. Andjel, Th. Mountford, and D. Valesin, Equilibrium of the interface of the grass- bushes-trees process, Bernoulli 24 (2018), no. 3, 2256–2277.

– R. Bissacot, E.O. Endo, A.C.D. van Enter, B. Kimura, and W.M. Ruszel, Contour methods for long-range ising models: Weakening nearest-neighbor interactions and adding decaying fields, Annales Henri Poincare´ 19 (2018), no. 8, 2557–2574.

– R. Bissacot, E.O. Endo, A.C.D. van Enter, and A. Le Ny, Entropic repulsion and lack of the g-measure property for dyson models, Commun. Math. Phys. 363 (2018), no. 3, 767–788.

– H.W. Broer, H. Hanßmann, and F. Wagener, Persistence properties of normally hyper- bolic tori, Reg. Chaot. Dyn. 23 (2018), no. 2, 212–225.

– X. Cai, K. Efstathiou, X. Xie, Y. Wu, Y. Shi, and L. Yu, A study of the effect of doughnut chart parameters on proportion estimation accuracy, Computer Graphics Forum (2018).

– L. Coquille, A.C.D. van Enter, A. Le Ny, and W.M. Ruszel, Absence of dobrushin states for 2d long-range ising models, J. Stat. Phys. 172 (2018), no. 5, 1210–1222.

– H. de Snoo and H. Woracek, The krein formula in almost pontryagin spaces. a proof via orthogonal coupling, Indagationes Mathematicae 29 (2018), no. 2, 714 – 729.

– H.S.V. de Snoo and H. Woracek, Compressed resolvents, q-functions and h0-resolvents in almost pontryagin spaces, Operator Theory: Adv. Appl. 263 (2018), 425–483.

– H. R. Dullin and H. Waalkens, Defect in the joint spectrum of hydrogen due to mon- odromy, Phys. Rev. Lett. (Editors’ Suggestion) 120 (2018), 020507.

54 Bernoulli Institute Annual Report

– H. Duminil-Copin, A.C.D. van Enter, and T. Hulshof, Higher order corrections for anisotropic bootstrap percolation, Probab. Theory Relat. Fields 172 (2018), no. 1–2, 191–243.

– J. Gao and K. Efstathiou, Self-consistent method and steady states of second-order oscillators, Phys. Rev. E 98 (2018), 042201.

– S. Hassi, H.S.V. de Snoo, and H. Winkler, Limit properties of eigenvalues in spectral gaps, Operator Theory: Adv. Appl. 263 (2018), 335–355.

– S. Hassi, M. Moller,¨ and H. de Snoo, Limit-point/limit-circle classification for hain–lust¨ type equations, Mathematische Nachrichten 291 (2018), no. 4, 652–668.

– S. Hassi, Z. Sebestyen,´ and H. de Snoo, Lebesgue type decompositions for linear relations and ando’s uniqueness criterion, Acta Sci. Math. (Szeged) 84 (2018), 465—- 507.

– M. Heydenreich, C. Hirsch, and D. Valesin, Uniformity of hitting times of the contact process, Latin American Journal of Probability and Statistics, Alea 15 (2018), 233–245.

– D.L. van Kekem and A.E. Sterk, Travelling waves and their bifurcations in the Lorenz-96 model, Physica D 367 (2018), 38–60.

– V. Krajnak and H. Waalkens, The phase space geometry underlying roaming reaction dynamics, Journal of Mathematical Chemistry 56 (2018), 2341–2378.

– Jean-Christophe Mourrat and Daniel Valesin, Spatial gibbs random graphs, Ann. Appl. Probab. 28 (2018), no. 2, 751–789.

– E.J. Pap, D. Boer, and H. Waalkens, Non-abelian nature of systems with multiple exceptional points, Phys. Rev. A 98 (2018), 023818.

– D. Prandi, L. Rizzi, and M. Seri, Quantum confinement on non-complete riemannian manifolds, Journal of Spectral Theory 8 (2018), no. 4, 1221–1280.

– A.E. Sterk and M.P. Holland, Extreme value laws and mean squared error growth in dynamical systems, Dynamics and Statistics of the Climate System: An Interdisciplinary Journal 3 (2018), 1–25.

– F. Tajik, M. Sedighi, A. Masoudi, H. Waalkens, and G. Palasantzas, Dependence of chaotic actuation dynamics of Casimir oscillators on optical properties and electrostatic effects, EPJ B 91 (2018), 71.

55 Bernoulli Institute Annual Report

– F. Tajik, M. Sedighi, A. Masoudi, H. Waalkens, and G. Palasantzas, Dependence of chaotic behavior on optical properties and electrostatic effects in double beam torsional casimir actuation, Phys. Rev. E 98 (2018), no. 2, 022210. – D.L. van Kekem and A.E. Sterk, Wave propagation in the Lorenz-96 model, Nonlinear Processes in Geophysics 25 (2018), no. 2, 301–314.

Book chapters

– Aernout C. D. van Enter, Scaling and inverse scaling in anisotropic bootstrap percola- tion, Probabilistic Cellular Automata (Theory, Applications and Future Perspectives) (Pierre-Yves LouisFrancesca R. Nardi, ed.), Springer, 2018, pp. 69–77.

3.5 External funding and collaboration

External funding

PhD grant (supervisor H.W. Broer, co-supervisor K. Efstathiou): Drs J. Gao has a China Scholarship Council grant for a PhD position, and has started working in Groningen on 1 October 2016.

PhD grant (supervisor H. Waalkens, co-supervisor K. Efstathiou): Drs B. Lin has a China Scholarship Council grant for a PhD position, and has started working in Groningen on 16 September 2017.

PhD grant (supervisors R. Bissacot and A.C.D. van Enter, cosupervisor D. Valesin): Drs E.O. Endo is funded by FAPESP, Brazil.

A tenure track position in the field of mathematical physics funded by the mathematics cluster GQT has been offered to Marcello Seri (Reading University). Seri is going to start in Groningen in June 2018. The funding which amounts to 345.000 Euro also covers a PhD position which will be filled soon after Seri’s arrival.

External collaboration

E. Verbitskiy collaborates with the graduate program Math-for-Industry, Kyushu University, Japan.

56 Bernoulli Institute Annual Report

A.C.D. van Enter collaborates with J. Miekisz and S. Taati, supported by the Polish Nationa Research center in the project Mathematical Models of Quasicrystals.

3.6 Further information

H.W. Broer is a member of the Royal Academy of Arts and Sciences (KNAW), afdeling Natuurkunde, sectie Wiskunde. He has given a talk at the 1st International Symposium on Mechanics in Aberdeen in July 2018.

K. Efstathiou gave invited talks at the conference Geometric Aspects of Momentum Maps and Integrability (CSF Ascona, 8-13 April 2018), at the IBS-CGP Workshop on Integrable Systems and Applications (Pohang, South Korea, 2-4 May 2018), and at the conference Symmetry and Perturbation Theory 2018 (SPT 2018; 3-10 June 2018). He gave seminar talks at the Analysis and Geometry Seminar at the University of Antwerp, and at the UvA-VU Dynamic Analysis Seminar. He has organized a special session on Geometry and Dynamics at the AIMS 2018 (Taipei, Taiwan, 5-9 July 2018), and a mini-workshop on integrable Hamiltonian systems at the University of Groningen on the occasion of the PhD defense of N. Martynchuk (20 September 2018).

A.C.D. van Enter gave talks at two workshops in Bochum, at a workshop at CIRM, Marseille, an ICM satellite conference in Sao Paulo, a minicourse in Bedlewo (Poland) and a colloquium talk at NYU Shanghai during a research visit. He also gave a farewell lecture in Groningen, on the occasion of which a symposium day was organised by G. Vegter and D. Valesin.

Van Enter is editor of Markov Proc. Rel. Fields , Braz. J. Prob. Stat., J. Stat. Phys., Stoch.Proc. Appl. and of Het Nederlands tijdschrift voor Natuurkunde. He is member of the steering committee of the Eurandom RSS programme. He was a member of the committee awarding the Stieltjesprize for the best thesis of 2017 in mathematics in the Netherlands.

H.S.V. de Snoo visited the University of Vaasa (Seppo Hassi) and several times the Technische Universitat¨ Graz (Jussi Behrndt); was guest professor in Graz in the Fall.

A.E. Sterk gave seminar talks at the VU/UvA Dynamics Seminar, the NWO Mathematics of Planet Earth meeting, and the Lorentz Center. In addition, he gave lectures in outreach activities aimed at prospective students and secondary school pupils taking the “Wiskunde D” curriculum. D. Valesin gave invited talks at a CIMPA School on Geometry and scaling of random structures (Buenos Aires, Argentina), at the Rhein-Mainz Kolloqium Stochastik (Mainz, Germany) and

57 Bernoulli Institute Annual Report at a Mini-workshop on stochastic population models on networks (Munich, Germany). He also spoke at Probability Seminars at Aix-Marseille Universite´ and at Universit’at´ Leipzig.

G. Vegter co-organized the Eigth Quantum Universe Symposium (March 28-29, 2018, Univer- sity of Groningen, The Netherlands). He is local coordinator of the INRIA-funded Associate Team Astonishing, consisting of groups of INRIA Nancy, U. Luxembourg, U. Marne-la-Vallee,´ Paris and RUG, and visited the INRIA-partner in Nancy for three weeks in the fall of 2018. He has been a member of the board of the JBI until the merger of this institute with ALICE (June 1, 2018), and he is a member of the Steering Committee of the Research theme Data Science and Systems Complexity of the FSE, and of the Steering Committee of the Research Priority Fundamentals of the Universe of the FSE.

Verbitskiy is member of the management team of the national mathematical research cluster ’Stochastics - Theoretical and Applied Research’ (STAR) (chairman, since 01/01/2018). He is an editor of the Pacific Journal of Mathematics for Industry, member of the editorial board of Springer series Mathematics for Industry, and member of the editorial board of Zeitschrift fur¨ Analysis und ihre Anwendungen.

H. Waalkens is member of the steering committee of the mathematics cluster NDNS+. He has coorganized the workshop Geometry of Chemical Reaction Dynamics in Gas and Condensed Phases in Telluride, Colorado (17 July - 27 July 2018; with S. Berry, J. Green and H. Teramoto) and the Conference on Dynamical Systems and Mechanics at the Chern Institute, Nankai University, Tianjin, China (10-14 September 2018; with U. Frauenfelder and L. Zhao). He has given invited talks at the conference Dynamics Days Europe in Loughborough in September 2018 and the conference AIMS 2018 in Taipei, Taiwan, in July 2018.

58 Bernoulli Institute Annual Report

4 Probability and Statistics

Group leader: Prof.dr. E.C. Wit

Tenured staff (BI members) source fte Prof. dr. E.C. Wit RuG till July 0.5 Prof. dr. T. Muller¨ RuG 1.0 Prof. dr. E.R. van den Heuvel RuG 0.0 Dr. W.P. Krijnen RuG 0.5

Tenure track source fte Dr. M.A. Grzegorczyk RuG 1.0

(Fellow) Postdoc source fte Dr. R. Jacobs RuG (as of October 1.0 first)

Emeritus source fte Prof. dr. W. Schaafsma RuG 0.0

PhD students S. Balafas RuG 1.0 (supervisor: Wit, Grzegorczyk) V.A. Bernal DSSC 1.0 (supervisor: Grzegorczyk, Horvatovich, Guryev) L. del Core San Raffaele, Milaan 0.2 (supervisor: Wit, Grzegorczyk) B. Hansen RuG 1.0 (supervisor: Muller)¨ F. Richter-Mendoza NWO 1.0 (supervisor: Wit) A. Salam Scholar 1.0 M. Schepers RuG 1.0 (supervisor: Muller)¨

59 Bernoulli Institute Annual Report

Guests H. Pazira, guestresearcher, Iran R. Uzupyte M. Bradonjic I. Artico Prof.dr. A. Dukkipatti A. Ahmadi Yazdi M. Arabpour D. Mitsche

60 Bernoulli Institute Annual Report

4.1 Research program

The focus of the Probability and Statistics group is network modelling, including random graphs and percolation, and inference. Our group is focused on developing modelling and inference approaches for this emerging field. The Statistics and Probability research pro- gramme spans the wide range of methodological developments and applied projects, from random graph models, sparse network model inference and systems biology, high-dimensional inference and inference of ODEs and SDEs.

4.1.1 Random graphs, percolation, combinatorics

Continuum percolation and random geometric graphs. One of the central questions in the theory of percolation is on the appearance of “giant” connected components. Ongoing work studies what exactly happens exactly at the “critical” choice of parameters for various models.

Hyperbolic models. It turns out that when percolation models or random graphs models are built over a space with a hyperbolic as opposed to a traditional (Euclidean) geometric space, spectacularly different types of behaviours appear to occur. The group studies a number of different such models, and various aspects of them.

Logical aspects of random graphs. Viewing random graphs through a logical lense is a topic that goes back several decades. It is arguably one of the most attractive aspects of modern random graph theory, with connection to various areas of modern computer science and combinatorics as well as more far removed branches of mathematics such as number theory and topology. The group in particular studies logical aspects of harder-to-analyse random graph models such as random planar graphs.

Topological combinatorics The group studies applications of methods from (algebraic) topology to combinatorial problems. And, loosely related to this, we also study so called “random simplicial complexes” which form a natural generalisation of random (geometric) graphs that form the theoretical basis for Topological Data Analysis.

61 Bernoulli Institute Annual Report

4.1.2 Network modelling and inference

From systems biology, where inference of large gene regulatory networks constitutes an important theme, to modern quantitative sociology with its interest in large social networks, novel methods to infer networks are required.

Social network inference. For social scientists discovering the driving forces behind social networks are of crucial importance. This requires on the one hand stochastic models enriched with covariates and on the other hand efficient inference methods to estimate the influence of these potential driving forces. We have developed novel penalized approaches to deal with the potential high-dimensional nature of the interactions. An example of such penalized social network modelling is shown in Figure 2, where social interactions between individual parliamentarians are partially explained by the political affiliation.

Dynamic Bayesian networks. The regulatory interactions in many cellular networks un- dergo temporal changes. The traditional (homogeneous) dynamic Bayesian network ap- proaches then fail to infer the right network structures from data. The Wit-Grzegorczyk group has pursued various efforts to relax the traditional homogeneity assumptions by combining dynamic Bayesian networks with multiple changepoint processes. The changepoints are used to divide the time series into segments with segment-specific network parameters. The new advanced models are referred to as non-homogeneous dynamic Bayesian networks.

Semi-mechanistic models. Semi-mechanistic models make use of the concept of gradient matching and translate ODE and SDE based models into non-linear Bayesian regression models. With semi-mechanistic models the regulatory network interactions can be modelled much more faithfully than with generic approaches, such as dynamic Bayesian network models. Unlike ODE and SDE models, semi-mechanistic models allow for model averaging, as all possible network structures can be scored in light of the data.

Penalized graphical models. One of the major goals of our research is therefore to address those over-fitting problems without losing too much modelling flexibility. To this end, our research focus is on developing novel regularized statistical methodologies and information- coupling schemes, which infer the right trade-off between flexibility and inference accuracy automatically from the available data. Our advanced methods offer maximal flexibility along with various possibilities for learning and tuning coupling strengths in light of the data, such

62 Bernoulli Institute Annual Report

77

75

79 ORG

74 18 60 25 47 34 TRAIN 19

51 ORG ORG 3 15 TRAIN 76 52 TRAIN 71 10 70 26 32 45 MEET ORG 5 13 6 55 62 31 MEET 28 20 67 46 40 27 MEET OPER 66 49 4 14 OPER ORG LOGST 33 OPER 2 1 LOGST ORG 39 38 23 42 MEET MEET 50 29 MEET 12 MEET 65 68 11 56 61 LOGST FIN 54 TRAIN 78 21 7 MEET LOGST ORG 41 24 OPERMEET 69 TRAIN 57 TRAIN 64 73 MEET 44 LOGST 63 ORG 72 MEET MEET 43 30 FIN 35 TRAIN 17 TRAIN LOGST OPER 36 TRAIN TRAIN TRAIN 9 22 LOGST 16 37 48 59

53 58

8

Figure 2: Ranciatti and Wit worked on modelling overlapping cluster: they applied their method to the Noordin-Top terrorist network, showing that the underlying social structure is much better understood as overlapping groups of terrorist rather than disjoint operatives.

that features which are not supported by the data are automatically down-rated. This maximizes inferential certainty for better-fitting model features.

ODE and SDE inference. Differential equations are the bread and butter of many quan- titative sciences, from climate studies to genomics. Often the aim is to match systems of stochastic (SDE) or ordinary differential equations (ODE) with data observed on the real-life process. Two important questions are (i) how to match the SDE or ODE to the data by varying various parameters and (ii) how to choose between alternative descriptions of the system based on the data.

63 Bernoulli Institute Annual Report

4.1.3 High-dimensional inference

Traditional statistical model, such as linear regression, considered a small number of covariates relative to the number of observations. With the advent of high-throughput and sensing technology, in many areas the data situation has reversed. Although the typical number of independent observations has not changed, the potential number of features has exploded. The field of high-dimensional inference is concerned with discovering and estimating the effect of true features that are hidden as needles in a haystack. Together with collaborators from Palermo, Wit has worked on differential geometric extensions of the least angle regression method for generalized linear models.

4.2 Overview of scientific results

In 2018, the research group has obtained the following scientific results:

Hyperbolic random graphs: Fountoulakis, Mitsche, Muller¨ and Schepers studied the tran- sition where the KPKVB random graph (often called “hyperbolic random graph”) becomes Hamiltonian, settling open problem in the literature.

Cheegers cuts in random point clouds: Muller¨ and Penrose closed a gap in the theory of optimal Cheeger cuts for geometric random graphs. This has relevance for the performance of certain kinds of machine learning algorithms.

Dynamic Bayesian networks For sparse data, non-homogeneous dynamic Bayesian net- works are often subject to inflated inference uncertainties. Grzegorczyk and Shafiee Kamalabad developed various advanced dynamic Bayesian network models with hierarchical coupling mechanisms. The improved models encourage the segment-specific interaction parameters to stay similar among time-segments. This can lead to significantly improved network re- construction accuracies. Grzegorczyk and Shafiee Kamalabad have submitted two papers on non-homogeneous dynamic Bayesian networks. One paper has meanwhile been published in Statistica Neerlandica. mTOR pathway. Grzegorczyk started a new collaboration with Prof. Kathrin Thedieck (Faculty of Medical Sciences) on reconstructing the mammalian target of rapomycin (mTOR)

64 Bernoulli Institute Annual Report signaling pathway using Bayesian networks and semi-mechanistic modelling approaches. One paper has meanwhile been accepted and will appear in Bioinformatics During 2018 X PhD students graduated and obtained their PhD degree.

4.3 Research subjects

S. Balafas (PhD): Sparse multivariate dynamic models V. Bernal (PhD): Clinical Big Data for multifactorial diseases R. Conijn (PD): Percolation, random graphs L. del Core (PhD): Statistical analysis of gene therapy data. M. Grzegorczyk (Ass.Prof): Dynamic Bayesian networks and Bayesian modelling B. Hansen (PhD): Percolation, random graphs R. Jacobs (PD): Bayesian variable selection methods for multilevel data in epidemiological and health applications M. Mahmoudi (PhD): Causal inference, ODE inference T. Muller¨ (Prof): Random graphs, percolation, discrete geometry, graph theory H. Pazira (PhD): Differential geometric least angle regression D. Pellin (PhD): Sparse stochastic differential equation modelling S. Ranciati (PhD): Bayesian inference for hierarchical mixture model F. Richter (PhD): Inference of species diversification models A. Salam (PhD): Advanced dynamic Bayesian networks M. Schepers (PhD): Hyperbolic random geometric graphs M. Shafiee (PhD): Dynamic Bayesian network inference M. Signorelli (PhD): Inference of stochastic block models E.C. Wit (Prof): Network inference, high-dimensional inference, Biostatistics

4.4 Publications

Dissertations

Pariya Behrouzi, Extensions of graphical models with applications in genetics and • genomics. 19-01-2018, E.C. Wit

Articles in scientific journals

Bernal Arzola, V., Bischoff, R., Guryev, V., Grzegorczyk, M. and Horvatovich, P. Exact • hypothesis testing for shrinkage based Gaussian Graphical Models. Bioinformatics, accepted and to appear (2018).

65 Bernoulli Institute Annual Report

Bernal Arzola, V., Guryev, V., Bischoff, R., Horvatovich, P., and Grzegorczyk, M. • Significance tests for Gaussian graphical models based on shrunken densities. In Proceedings of the 33rd Inter- national Workshop on Statistical Modelling (IWSM), University of Bristol, UK, 16-20 July 2018 (7 2018), vol. 2, p. 27-30.

Grzegorczyk, M. Editorial introduction. Statistica Neerlandica 72, 3 (8 2018), 178–178. • Gudmundson, J., Kotsitsyna, I., Loffler,¨ M., Muller,¨ T., Sacristan,´ V., and Silveiro, • R. Theoretical analysis of beaconless geocast protocols in 1d. In Proceedings of the Fifteenth Workshop on Analytic Algorithmics and Combinatorics (ANALCO) (2018), M. Nebel and S. Wagner, Eds., SIAM, p. 6276.

Heinig, P., Muller,¨ T., Noy, M., and Taraz, A. Logical limit laws for minor-closed classes • of graphs. Journal of Combinatorial Theory, Series B 130 (5 2018), 158–206. 4

Shafiee Kamalabad, M., and Grzegorczyk, M. Improving nonhomogeneous dynamic • bayesian networks with sequentially coupled parameters. Statistica Neerlandica 72, 3 (8 2018), 281–305.

Litsios, A., Ortega, Wit, E., and Heinemann, M. Metabolic-flux dependent regulation of • microbial physiology. Current Opinion in Microbiology 42 (4 2018), 71–78. Copyright 2017 The Authors. Published by Elsevier Ltd.. All rights reserved.

Mahmoudi, S., and Wit, E. Estimating causal effects from nonparanormal observational • data. The international journal of biostatistics 14, 2 (12 2018).

Muller,¨ T., and Noy, M. The first order convergence law fails for random perfect graphs. • Random structures & algorithms 53, 4 (12 2018), 717–727.

Pazira, H., Augugliaro, L., and Wit, E. Extended differential geometric lars for high- • dimensional glms with general dispersion parameter. Statistics and Computing 28, 4 (7 2018), 753–774.

Shafiee Kamalabad, M. and Grzegorczyk, M. Non-homogeneous dynamic Bayesian • networks with edge-wise coupled parameters. In Proceedings of the International Workshop on Statistical Modelling (IWSM), University of Bristol, UK, 16-20 July 2018 (7 2018), vol. 1, p. 270-275.

Shafiee Kamalabad, M., Heberle, A., Thedieck, K., and Grzegorczyk, M. Partially • non-homogeneous dynamic Bayesian networks based on Bayesian regression models with partitioned design matrices. Bioinformatics, accepted and to appear (2018).

66 Bernoulli Institute Annual Report

Stiekema, A., Islam, M., Liemburg, E., Castelein, S., van den Heuvel, E., van Weeghel, • J., Aleman, A., Bruggeman, R., der Meer, L., and GROUP Investigators. Long-term course of negative symptom subdomains and relationship with outcome in patients with a psychotic disorder. Schizophrenia Research 193 (3 2018), 173–181. Copyright 2017 Elsevier B.V. All rights reserved. Wit, E. A penalized inference approach to stochastic block modelling of community • structure in the italian parliament. Journal of the Royal Statistical Society. Series C: Applied Statistics 67, 2 (2 2018), 355–369. Wit, E., Augugliaro, L., Pazira, H., Gonzalez,¨ J., and Abegaz, F. Sparse relative risk • regres- sion models. Biostatistics (2018). Zhan, Z., de Bock, G., and van den Heuvel, E. Optimal unidirectional switch designs. • Statistics in Medicine 37, 25 (11 2018), 3573–3588.

4.5 External funding and collaboration

4.5.1 External funding

During 2018, the group has been involved in the following successful grant applications:

NWO : visitor grant to host Prof. A. Dukkipatti (IISc Bangalore) • The members of the group are supported in a number of ways. The PhD students Hansen and Schepers and postdoc Basu are supported throught an NWO Vidi grant, and the postdoc Taati is supported through an NWO TOP C-2 grant. Francisco Richter is funded by a NWO grant (Mathematics for Planet Earth). Nazia Gill and Mahdi Mahmoudi are funded by their home governments in Pakistan and Iran, respectively. Danilo Pellin, Saverio Ranciati and Mirko Signorelli are funded by their home institutions in Italy (San Raffaele Milano, Bologna and Padova, respectively) and the Statistics and Probability unit for conference travel. S. Balafas, Pariya Behrouzi and Mahdi Shafiee are funded by the RuG. Victor Bernal is funded by a DSSC grant. Hassan Pazira is self-funded. Abdul Salam is funded by Malakand University (Malakand, Pakistan). Luca Del Core is funded by SR-TIGET (Milan, Italy).

4.5.2 External collaboration

Muller’s¨ ongoing collaborations include: a collaboration with N. Fountoulakis (Birm- • ingham U.), P. van der Hoorn (Boston U.) on hyperbolic random graphs, with M. Stehlik

67 Bernoulli Institute Annual Report

(G-SCOP, Grenoble) on topological methods in combinatorics; with M. Noy (UPC Barcelona) on logical aspects of random graphs; with N. Wormald (Monash U.) on random lifts; with M. Loffler¨ (Utrecht U.) on geocast protocols; with E. Broman and J. Tykesson (Gothenburg U.) on hyperbolic continuum percolation; with S. Haber (Bar Ilan U.) on logical aspects of random graphs; with F. Skerman on random premutations. Grzegorczyk is collaborating with Dirk Husmeier (Glasgow, UK) and is collaborating • with researchers from SR-TIGET (Milan, Italy).

Wit is leading the EU-funded Cooperation in Science and Technology (COST) project • on statistical network science. The collaboration involves some 350 people across 33 European countries and will be funded for 4 years. It started in June 2016. Wit is collaborating with L. Augugliaro and A. Abbruzzo (University of Palermo), C. • Viroli (University of Bologna), M.C. di Serio (University of San Raffaele Milano) and V. Vinciotti (Brunel University).

4.6 Further information

Grzegorczyk is an associate editor of the journals: Computational Statistics and the Journal of Applied Statistics, and he was guest editor of a special issue on Statistical Modelling of the journal Statistica Neerlandica. He is a member of the representative council of the International Biometrics Society (Dutch region), and has been elected into the Executive Committee of the Statistical Modelling Society for the period 2019-2020.

Muller¨ was on the programme committees for Workshop on Algorithms and Models for the Web Graph (WAW 2019); and co-organiser of the fourth “STAR Workshop on Random Graphs”. He was invited on the editorial board of the journal Indagationes Mathematicae, and to become a panel member for the Flemish Science foundation (FWO). Muller¨ was invited to speak at the Oberwolfach (MFO) mini-workshop on “Positional Games”, the workshop on “Large Networks and Random Graphs” in Frankfurt, a mini-workshop on Graphs in Nijmegen and the third workshop on “Critical and Collective Effects in Graphs and Networks”, in Eindhoven.

Wit is an associate editor of Statistical Applications in Genetics and Molecular Biology and Biometrics. He is member of the Board of Directors of the International Biometrics Society. He is president of the Dutch Biostatistics Society (BMS-ANed).

68 Bernoulli Institute Annual Report

5 Systems, Control and Applied Analysis

Group leader: Prof.dr. A.J. van der Schaft

Tenured staff (BI members) source fte Prof.dr. M.K. Camlibel RuG 1.0 Prof.dr. H.L. Trentelman RuG 1.0 Prof.dr. A.J. van der Schaft RuG 0.6 Prof.dr. S. Trenn RuG 1.0 Dr. A. Waters RuG 1.0

Tenure Track source fte Dr. B. Besselink RuG 1.0

PhD students A.M. Burohman (since March) DSSC 1.0 (supervisors: M.K. Camlibel, J. Scherpen (ENTEG)) J. Eising RUG 1.0 (supervisors: M.K. Camlibel) S. Hossain RuG 1.0 (supervisors: S. Trenn, M.K. Camlibel) M. Jeeninga NWO 0.5 (supervisors: C. De Persis (ENTEG), A.J. van der Schaft ) J. Jia CSC 1.0 (supervisors: M.K. Camlibel, H.L. Trentelman) J. Jiao CSC 1.0 (supervisors: M.K. Camlibel, H.L. Trentelman) M. Jozsa (till September) BI/Armines 0.5 (supervisors: M.K. Camlibel, M. Petreczky (Lille)) F. Koerts (till October) NWO 1.0 (supervisors: A.J. van der Schaft, C. De Persis (ENTEG)) P. Monshizadeh (till April) STW 1.0 (supervisors: A.J. van der Schaft, C. De Persis (ENTEG)) R. Reyes Baez CONACYT 1.0 (supervisors: A.J. van der Schaft, B. Jayawardhana (EN- TEG)

69 Bernoulli Institute Annual Report

PhD students K. Shomalzadeh (since May) RuG 1.0 (supervisors: M.K. Camlibel, J. Scherpen (ENTEG)) T. Stegink (till August 15) NWO 0.5 (supervisors: A.J. van der Schaft, C. De Persis (ENTEG)) H. van Waarde (from December) DSSC 0.5 (supervisors: M.K. Camlibel, P. Tesi (ENTEG)) C. Wang (since April first) CSC 1.0 (supervisors: H.L. Trentelman, M.K. Camlibel) L. Wang Ubbo Emmius 1.0 (till February) Sandwich (supervisors: A.J. van der Schaft, B. Maschke (Lyon)) P. Wijnbergen (since August first) NWO 1.0 (supervisor: S. Trenn) T. Xu (since December) RUG 1.0 (supervisors: A. Waters, S. Trenn)

Guests B. Anderson, Australian National University Dr. S Trenn, University of Kaiserslautern, Germany I. Oner, Gebze Technical University, Turkey J. Trumpf, Australian National University K. Fujimoto, Kyoto University M. Cucuzzella, PhD student, University of Pavia, Italy M. Farokhian, PhD student, Shiraz University of Technology, Iran

70 Bernoulli Institute Annual Report

5.1 Research Program

The research program Systems, Control and Applied Analysis (SCAA) is devoted to the analysis, control and optimization of complex large-scale open dynamical systems evolving in time. The dynamics is described by ordinary or partial differential equations, or is a mixture of continuous and discrete dynamics. This dynamical behavior is not only sought to be analyzed, but to be influenced (’controlled’) and optimized as well; by the addition of feedback loops, and by the interconnection to other dynamical systems (controller design). Furthermore, dynamical data are used to identify the underlying model, or to approximate it by a model of lower dimension. Typically, the system models under consideration are described by under-determined sets of equations. As a result, there are free variables in the system description (corresponding to ’inputs’), which together with information about the current state of the system (corresponding to ’outputs’) model the interaction with other systems or the environment. Furthermore, the systems point of view is emphasized, in the sense that large-scale dynamical systems are viewed as networks of interconnected systems, where the overall dynamics is determined by the dynamics of the system components plus the network and feedback structure. This point of view is prevailing in many areas of engineering science and economics, and is receiving increasing attention in the life sciences (’reverse engineering’). The modeling, control, and optimization of complex dynamics brings together a variety of mathematical theories and tools, from linear algebra, analysis, geometric control, multi-physics modeling and stability theory, to algebraic graph theory, Hamiltonian systems and (distributed) optimization. The members of the program have close collaboration and multiple joint projects with col- leagues working in other scientific disciplines. In particular, there is a close collaboration with the control engineering groups DTPA and SMS at the neighboring Engineering and Technology Institute (ENTEG), under the umbrella of the Jan C. Willems Center for Systems and Control. Although the focus in the program is on fundamental mathematical developments, the research is motivated by several application areas, and applied to these in collaboration with colleagues. Current application topics include energy systems (stability and control of power networks, systems integration, dynamic pricing), distribution networks, cardiac modeling, systems biology (chemical reaction networks, regulation), mechatronics, smart factories and intelligent transportation systems.

71 Bernoulli Institute Annual Report

Structure of the program

The main lines of research in the program are:

1. Network dynamics and control (Kanat Camlibel, Harry Trentelman, Arjan van der Schaft, Bart Besselink, Oleksandr Ivanov, Mark Jeeninga, Monika Jozsa, Jia Jiajia, Junjie Jiao, Pooya Monshizadeh, Li Wang, Henk van Waarde, Azka Burohman) Network dynamics and dynamical multi-agent systems arise in many fields of engi- neering and natural sciences. Systems and control theory contributes to this area by providing concepts and tools for the study of structural properties such as controllability (leader-follower networks), model reduction and identifiability, and for their control in- cluding synchronization and consensus dynamics. This entails a close interplay between geometric systems and control theory on the one hand, and algebraic graph theory on the other. Applications include power and sensor networks and dynamical distribution networks.

2. Geometric modeling and control of multi-physics systems (Arjan van der Schaft, Filip Koerts, Pooya Monshizadeh, Rodolfo Reyes Baez, Tjerk Stegink, Li Wang) Port-Hamiltonian systems constitute an extension of Hamiltonian systems where exter- nal interaction and energy-dissipating ports are taken into account, and the underlying geometry is derived from the interconnection structure of the complex system. The aim of this research is to provide a systematic geometric theory for the modeling, analysis and simulation of multi-physics, lumped- and distributed parameter, systems. Current focal themes are the geometric modeling and analysis of power systems and of ther- modynamic systems. The port-Hamiltonian formulation is employed for controller design, leading to physically inspired and robust control strategies. Applications include stabilization and demand-supply matching in power systems, control of robotic systems and distribution networks, and analysis and control of chemical reaction networks.

3. Mathematical systems theory (Kanat Camlibel, Harry Trentelman, Bart Besselink, Stephan Trenn, Arjan van der Schaft, Junjie Jiao, Monika Jozsa, Jaap Eising, Noorma Yulia Megawati, D. Kocoglu, Paul Wijnbergen, S. Hossain) Mathematical systems theory deals with the modeling and analysis of open and inter- connected systems. This naturally leads to models containing differential and algebraic equations, called DAE systems. Current research themes concern equivalence and mini- mality notions, and model reduction. Furthermore, physical systems often do not exhibit an a priori fixed information flow direction. In the behavioral approach, all external sys- tem variables are therefore in first instance treated on an equal footing. Hybrid systems are a mixture of interacting continuous and discrete dynamics, and arise naturally in

72 Bernoulli Institute Annual Report

embedded systems and physical systems modeling, including switched DAE systems and convex processes. Important research issues concern the analysis of solution tra- jectories, and the structural properties of controllability and stabilizability, as well as the design of controllers. The mathematical analysis of switched and piecewise-affine and systems is heavily intertwined with convex optimization theory and non-smooth analysis. Another line of research concerns the compositional analysis and design of interconnected systems, developing notions and tools of assume-guarantee reasoning and contract-based design using geometric control theory.

4. Modeling, control and optimization of energy systems (Arjan van der Schaft, Kanat Camlibel, Stephan Trenn, Mark Jeeninga, Filip Koerts, Pooya Monshizadeh, Tjerk Stegink, K. Shomalzadeh) Power networks, from high-voltage distribution networks to AC or DC micro-grids, constitute an application area of growing importance and interest. Furthermore, there is an increasing trend for integration with other energy systems such as gas distribution networks. This theme is concerned with the development of a sound mathematical framework for the modeling, optimization and control of large-scale energy systems. This includes the systematic modeling of components such as synchronous generators and converters, as well as of the transmission line network and the corresponding load flow equations. Based on these models fundamental problems of stability and power sharing are addressed, as well as optimal demand-supply matching by dynamic pricing, coupling physical dynamics to market dynamics. Furthermore, integration of power networks with other energy systems leads to large-scale distributed optimization problems. This research is mostly carried out in a collaborative effort with colleagues from the Engineering and Technology Institute Groningen (ENTEG).

5. Control of distributed-parameter systems and inverse problems (Alden Waters, Arjan van der Schaft, Stephan Trenn, Hugo Carillo, Teke Xu) This research theme is concerned with the analysis, control and estimation of systems described by partial differential equations. Current themes of interest are approximations to solutions for coupled systems of hyperbolic wave equations, and issues of short time well-posedness and parameter recovery from solution waves. Some current collaboration focuses on well-posedness of coupled systems of hyperbolic and switched DAE balance laws in the context of modeling blood flow.

5.2 Research subjects

Staff: B. Besselink: compositional analysis and control of large-scale interconnected systems, model reduction, analysis and control of cooperative transportation systems.

73 Bernoulli Institute Annual Report

M.K. Camlibel: analysis of piecewise affine dynamical systems, model reduction of switched linear systems, controllability of multi-agent systems, optimization of energy systems, nons- mooth optimization and control. A.J. van der Schaft: geometric network modeling, analysis and control of complex systems, nonlinear systems and control theory, network dynamics and model reduction, modeling and control of power networks and thermodynamic systems. S. Trenn: differential-algebraic equations, switched systems, observer design, funnel control, modeling of power grids H.L. Trentelman: network dynamics, model reduction of network dynamics, synchronization and controllability of networked systems, control in a behavioral setting. A.M.S. Waters: inverse problems and integral geometry, well-posedness of partial differential equations, coupled systems, low regularity wave equations, fluid flow in human heart struc- tures.

PhD students: Azka Burohman: model reduction for dynamical networks. H. Carillo: inverse problems and parameter estimation, pseudo-differential operators. J. Eising: analysis and optimal control of convex processes. K. Shomalzadeh: distributed optimization, systems integration. S. Hossain: model reduction, switched systems, differential-algebraic equations O. Ivanov: chemical reaction networks in systems biology. M. Jeeninga: load flow equations in power networks. Jia Jiajia: structural controllability of networked systems. Junjie Jiao: distributed linear quadratic optimal control of networked systems. M. Jozsa: input-output properties of interconnected systems, network structure and control and estimation. F. Koerts: optimization of dynamical models of power networks. D. Kocoglu (TU Kaiserslautern): PDE networks, differential-algebraic equations, distribu- tional solutions N.Y. Megawati: bisimulation of linear DAE systems and control by interconnection P. Monshizadeh: analysis of micro-grids, frequency and voltage regulation. R. Reyes Baez: port-Hamiltonian systems and incremental passivity. T.W. Stegink: power networks, optimization and dynamic pricing methods. H. van Waarde: network identification, synchronization. Li Wang: chemical reaction networks and thermodynamics. P. Wijnbergen: optimal control, switched systems, differential-algrbaic equations

74 Bernoulli Institute Annual Report

5.3 Publications

Dissertations

A. Ivanov, Network design and the response of a system to stress and deterioration. • Faculty of Science and Engineering, promotores: F.J. Weissing (Gelifes), A.J. van der Schaft, Faculty of Science and Engineering, University of Groningen, March 2018. R. Kausar, Analysis and modeling of water distribution network in the framework of • switched DAEs, promotor S. Trenn, TU Kaiserslautern, Germany, September 2018. F. Kusters,¨ Switch observability for differential-algebraic systems, promotor S. Trenn, • TU Kaiserslautern/ITWM, Germany, March 2018. P. Monshizadeh, Modeling and Control of Power Systems in Microgrids, promotores A.J. • van der Schaft, C. De Persis (ENTEG), Faculty of Science and Engineering, University of Groningen, September 2018. T.W. Stegink, Energy-based analysis and control of power networks and markets, pro- • motores C. De Persis (ENTEG), A.J. van der Schaft, Faculty of Science and Engineering, University of Groningen, December 2018. L. Wang, Modeling and Analysis of Non-Isothermal Chemical Reaction Networks, • promotores A.J. van der Schaft, B.M. Maschke (Univ. Lyon-1), Faculty of Science and Engineering, University of Groningen, April 2018.

Articles in scientific journals

B. Besselink and S. Knorn. Scalable input-to-state stability for performance analysis of • large-scale networks. IEEE Control Systems Letters, 2(3):507–512, 2018. M. Cucuzzella, S. Trip, C. De Persis, X. Cheng, A. Ferrara, and A.J. van der Schaft. A • robust consensus algorithm for current sharing and voltage regulation in DC microgrids. IEEE Transactions on Control Systems Technology, 2, pp. 1583-1595, 2018. PAGES? Fan Zhang, H.L. Trentelman, Gang Feng, and J.M.A. Scherpen. Absolute stabilization • of Lur’e systems via dynamic output feedback, European Journal of Control, Vol. 44, pp. 15 – 26, 2018. H.R. Feyzmahdavian, B. Besselink, and M. Johansson. Stability analysis of monotone • systems via max-separable Lyapunov functions. IEEE Transactions on Automatic Control, 63(3), pp. 643–656, 2018.

75 Bernoulli Institute Annual Report

H.J. Jongsma, P. Mlinaric, S. Grundel, P. Benner, and H.L. Trentelman. Model reduction • of linear multi-agent systems by clustering with H-2 and H-infinity error bounds, Mathematics of Control, Signals and Systems, 30 (6), 2018.

H.J. Jongsma, H.L. Trentelman, and M.K. Camlibel. Model reduction of networked • multi-agent systems by cycle removal, IEEE Transactions on Automatic Control, Vol.63, No. 3, pp. 657 - 671, 2018.

S.-Z. Khong and A.J. van der Schaft. The converse of the passivity and small-gain • theorems for input-output maps. Automatica, 97, pp.58–63, 2018.

F. Kusters¨ and S. Trenn. Switch observability for switched linear systems. Automatica, • 87, 121–127, 2018.

N.Y. Megawati and A.J. van der Schaft. Bisimulation equivalence of differential- • algebraic systems. Int. Journal of Control, 91(1), pp. 45–56, 2018.

N. Monshizadeh, P. Monshizadeh, R Ortega, and A.J. van der Schaft. Conditions on • shifted passivity of port-Hamiltonian systems. Systems & Control Letters, 123, pp. 55-61, 2018.

P. Monshizadeh, N. Monshizadeh, C. De Persis, and A.J. van der Schaft. Output • impedance diffusion into lossy power lines. IEEE Transactions on Power Systems, 2018

S. Trip, M. Cucuzzella, C. De Persis, A.J. van der Schaft, and A. Ferrara. Passivity • based design of sliding modes for optimal Load Frequency Control. IEEE Transactions on Control Systems Technology, 99, pp. 1–14, 2018.

A.J. van der Schaft and B. Maschke. Geometry of thermodynamic processes. Entropy, • 20 (12), pp. 925-947, 2018.

A.J. van der Schaft and B. Maschke. Generalized port-Hamiltonian DAE systems. • Systems & Control Letters, 121, pp. 31-37, 2018.

L. Wang, B. Maschke and A.J. van der Schaft. Port-Hamiltonian modeling of non- • isothermal chemical reaction networks. J. Math Chem, 56 (6), 1707–1727, 2018.

A. Waters and E. Merkurjev. Asymptotics for optimal design problems for the Schrodinger¨ • equation with a potential Journal of Optimisation, 16 p., 816 – 845, 2018.

J. Ilmavirta and A. Waters. Recovery of the sound speed for the acoustic wave equation • from phaseless measurements. Communications in mathematical sciences, 16(4), pp. 1017-1041, 2018.

76 Bernoulli Institute Annual Report

A. Waters. Recovery of attenuation coefficients from phaseless measurements for the • Helmholtz equation. Communications in mathematical sciences, 16(2), pp. 579-587, 2018.

Weixin Han, H.L. Trentelman, Zhenhua Wang, and Yi Shen. Towards a minimal order • distributed observer for linear systems, Systems and Control Letters, Vol. 114, pp. 59 - 65, 2018.

H.J. van Waarde, P. Tesi, and M.K. Camlibel. Identifiability of undirected dynamical • networks: A graph theoretic approach, IEEE Control Systems Letters, vol. 2, no. 4, pp. 683–688, 2018.

Articles in refereed conference proceedings

T. Gross, S. Trenn, and A. Wirsen. Switch induced instabilities for stable power system • DAE models, IFAC-PapersOnLine (Proc. IFAC Conf. Analysis Design Hybrid Systems (ADHS 2018)), pp. 127–132, 2018.

R. Kausar and S. Trenn. Water hammer modeling for water networks via hyperbolic • PDEs and switched DAEs, in C. Klingenberg, M. Westdickenberg (Ed.): Theory, Numerics and Applications of Hyperbolic Problems II (Conference proceedings HYP 2016), pp. 123-135, Springer, Cham, 2018.

M. Jeeninga, C. De Persis, and A.J. van der Schaft. Graph theoretic formulae for the • determinant and adjugate of matrices carrying graph structure, 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), August 27-28, 2018, Groningen, Netherlands, IFAC-PapersOnLine, pp. 259–264, 2018.

M. Jeeninga, A.J. van der Schaft, and C. De Persis. A scheme for computing the transfer • function of power grids with grounded capacitors, 23rd International Symposium on Mathematical Theory of Networks and Systems (MTNS2018), pp. 784–786, 2018.

J. Jia, H.L. Trentelman, W. Baar, and M.K. Camlibel. A sufficient condition for colored • strong structural controllability of networks, Proceedings of the 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), August 27-28, Groningen, The Netherlands, pp. 16 - 21, 2018.

Junjie Jiao, H.L. Trentelman, and M.K. Camlibel. A suboptimality approach to dis- • tributed H-2 optimal control, Proceedings of the 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), August 27-28, Groningen, The Netherlands, pp. 154 - 159, 2018.

77 Bernoulli Institute Annual Report

P. Kotyczka, B. Maschke, and A.J. van der Schaft. Port-Hamiltonian systems on graphs • and complexes, 23rd International Symposium on Mathematical Theory of Networks and Systems (MTNS2018), pp. 154–157, 2018.

B. Maschke and A.J. van der Schaft. Homogeneous Hamiltonian Control Systems Part • II: Application to thermodynamic systems. IFAC-PapersOnLine 51 (3), pp. 7–12, 2018.

S. Naderi Lordejani, B. Besselink, M.H. Abbasi, G.-O. Kaasa, W.H.A. Schilders, and • N. van de Wouw. Model order reduction for managed pressure drilling systems based on a model with local nonlinearities. In Proceedings of the 3rd IFAC Workshop on Automatic Control in Offshore Oil and Gas Production, Esbjerg, Denmark, 50–55, 2018.

R. Reyes-Baez, A.J. van der Schaft, and B. Jayawardhana. Passivity based distributed • tracking control of networked Euler-Lagrange systems, 7th IFAC Workshop on Dis- tributed Estimation and Control in Networked Systems (NecSys18), August 27-28, 2018, Groningen, Netherlands, IFAC-PapersOnLine 51(23), pp. 136–141, 2018.

T. Stegink, A. Cherukuri, C. De Persis, A.J. van der Schaft, and J. Cortes. Stable • interconnection of continuous-time price-bidding mechanisms with power network dynamics, Power Systems Computation Conference (PSCC), pp. 1-6, Dublin, Ireland, 2018.

T. Stegink, A. Cherukuri, C. De Persis, A. J. van der Schaft, and J. Cortes. Integrating • iterative bidding in electricity markets and frequency regulation Proceedings of the American Control Conference, Milwaukee, Wisconsin, USA, pp. 6182–6187, 2018.

A.J. van der Schaft and B. Maschke. A symplectic formulation of open thermodynamic • systems, 23rd International Symposium on Mathematical Theory of Networks and Systems (MTNS2018), pp. 405–407, 2018.

A.J. van der Schaft and B. Maschke. Homogeneous Hamiltonian Control Systems Part • I: Geometric Formulation. IFAC-PapersOnLine 51 (3), pp. 1–6, 2018.

Weixing Han, H.L. Trentelman, Zhenhua Wang, and Yi Shen. Distributed fault detection • observer design for linear systems, Proceedings of the 23rd International Symposium on the Mathematical Theory of Networks and Systems, MTNS 2018, July 16-20, Hong Kong, China, pp. 838 - 843, 2018.

I. Oner and M.K. Camlibel. On stabilizability of strict convex processes with arbitrary • stability domains, Proceedings of the 23rd International Symposium on the Mathemati- cal Theory of Networks and Systems, MTNS 2018, July 16-20, Hong Kong, China, pp. 266 - 269, 2018.

78 Bernoulli Institute Annual Report

H.J. van Waarde, P. Tesi, and M.K. Camlibel. Identifiability of undirected dynamical • networks: A graph theoretic approach, Proceedings of the 57th IEEE Conference on Decision and Control (CDC’18), December 17-19, Miami, USA, pp. 683–688, 2018. H.J. van Waarde, P. Tesi, and M.K. Camlibel. Topological conditions for identifiabil- • ity of dynamical networks with partial node measurements, 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), August 27-28, 2018, Groningen, Netherlands, IFAC-PapersOnLine 51(23), pp. 319–324, 2018.

5.4 Editorial activities

M.K. Camlibel: Associate Editor for the journal IEEE Transactions on Automatica Control Associate Editor for the journal SIAM Journal of Control and Opimization Associate Editor for the journal Systems & Control Letters Conference Associate Editor for the International Conference on Complex Networks and Their Applications, Cambridge, United Kingdom

S. Trenn: Member of the Editorial Board for the book series Differential-Algebraic Equations Forum. Associate Editor for the journal Systems & Control Letters.

H.L. Trentelman: Senior Editor for the journal IEEE Transactions on Automatic Control

A.J. van der Schaft: Member of the Editorial Board for the journal Annual Reviews in Control Member of the Editorial Board for the journal Journal of Geometric Mechanics.

5.5 Further signs of recognition and news items

Bart Besselink was member of the National Organizing Committee for the 7th IFAC • Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), Groningen. He was co-organizer for the EU-MORNET Groningen Autumn School on Model Order Reduction, Groningen. Kanat Camlibel is member of • IFAC Technical Committee of Discrete Event and Hybrid Systems;

79 Bernoulli Institute Annual Report

IFAC Technical Committee of Control Design; IFAC Technical Committee of Linear Systems; International Program Committee member for the International Conference on Complex Networks and Their Applications, Cambridge, United Kingdom; International Program Committee co-chair for the 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), Groningen. Harry Trentelman is a member of the IEEE Technical Committee on Behavioral Systems • and Control. He was an appointed member of the Board of Governors of the IEEE Control Systems Society, and a member of the advisory board of the Lorentz Center. He was a member of the National Organizing Committee for the 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), Groningen.

Arjan van der Schaft is member of the international Steering Committee of the Interna- • tional Symposium on Mathematical Theory of Networks and Systems (MTNS). He is member of the IFAC Technical Committee on Non-linear Control Systems; IFAC Technical Committee on Distributed Parameter Systems; IEEE Control Systems Society Technical Committee on Systems Biology. He is international Program Committee Member of the IFAC Symposium on Lagrangian and Hamiltonian Methods in Nonlinear Control (LHMNC), May 2018, and the 24th Int. Symp. Mathematical Theory of Networks and Systems (MTNS), August 2020, Cambridge. He was member of the National Organizing Committee for the 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys18), Groningen.

Alden Waters is co-organiser of a minisymposium on inverse problems for Equadiff, • Leiden, July 2019. She is current reviewer for MathSciNet (AMS reviews).

80 Bernoulli Institute Annual Report

5.6 Further information

List of seminars 2018:

1. December 6, Tjerk Stegink (University of Groningen) Energy-based analysis and control of power networks and markets 2. November 29, Dimos Dimarogonas (KTH, Sweden) Distributed hybrid control of multi-robot systems under spatiotemporal specifications 3. November 29, Danial M. Senejohnny (University of Groningen) Resilience of coordination networks: data availability and integrity 4. November 23, Shuai Feng (University of Groningen) Resilient control under denial-of-service attacks 5. November 2, Witold Respondek (INSA de Rouen Normandie, France) Linear DEA’s versus linear control systems 6. November 2, Witold Respondek (INSA de Rouen Normandie, France) Nonlinear DAE’s, nonlinear implicit control systems, and their linearization 7. November 4, Henrik Sandberg (KTH Royal Institute of Technology, Sweden) Thermodynamic costs in implementing Kalman-Bucy filters 8. October 31, Xiaodong Cheng (University of Groningen) Model reduction of network systems with structure preservation 9. September 26, Andre Teixeira (Uppsala University, Sweden) Sensititvity metrics for secure control systems 10. September 24, Pooya Monshizadeh (University of Groningen) Modeling and control of power systems in microgrids 11. September 24, Romeo Ortega (Centrale Supelec, France) New results on identification and adaptive control using dynamic regressor extension and mixed parameter estimation 12. September 5, Vincent Andrieu (University of Lyon, France) Expressing an observer in preferred coordinates by transforming an injective immersion into a surjective diffeomorphism 13. September 4, Achim Ilchmann (University of Ilmenau, Germany) About funnel control

81 Bernoulli Institute Annual Report

14. August 29, Hideaki Ishii (Tokyo Institute of Technology, Japan) Fault tolerant clock synchronization over unreliable channels in wireless sensor networks 15. July 13, Nirav Bhatt (IIT Madras, India) Identification of distribution network topology from steady-state data 16. July 10, Kalle Johansson (KTH, Stockholm, Sweden) Event-triggered feedback and feedforward control subject to actuator saturation 17. July 10, Giancarlo Ferrari Trecate (EPFL, Lausanne, Switzerland) Control and cybersecurity for microgrids with flexible structure 18. July 6, Erik Weitenberg (University of Groningen) Robustness of electrical power networks 19. June 29, George Weiss (Tel Aviv University, Israel) Synchronverters used for damping inter-area oscillations in two-area power systems 20. June 26, Francesco Vasca (University of Sannio, Italy) and Raffaele Iervolino (Univer- sity of Naples Federico II, Italy) A consensus policy and piecewise quadratic stability for heterogeneous opinion dynam- ics 21. May 22, Brian Anderson (Australian National University, Australia) Distance-based rigid formation control with signed area constraints 22. May 8, Hyungbo Shim (Seoul National University, South Korea) Blending vector fields by strong coupling for heterogeneous multi-agent systems and its applications 23. April 26, Paul Van den Hof (TU Eindhoven) Data-driven modeling in linear dynamic networks 24. April 23, Ashish Cherukuri (ETH, Zurich) Optimization in networked cyber-physical systems: distributed and data-driven methods 25. April 23, Andreas Kasis (University of Cambridge, UK) Distributed secondary frequency control schemes for stability and optimality in power networks 26. April 18, James Riehl (Washington University in St. Louis, USA) Coordinated solutions to complex optimization and control problems on networks 27. April 17, Sebastian Trip (University of Groningen) Modular design of flow networks: from practice to theory, and back

82 Bernoulli Institute Annual Report

28. April 16, Vahab Rostampour (TU Delft) Distributed data driven decision making with feasibility performance certificates

29. April 12, Dario Bauso (University of Sheffield) When controls meet economics and social sciences 30. April 6, Li Wang (University of Groningen) Modeling of non-isothermal chemical reaction networks

31. April 3, Bart Besselink (University of Groningen) Control and coordination of large-scale freight transportation systems 32. March 20, Rick Middleton (University of Newcastle, UK) Neuronal calcium stress: An in-silico model motivated by Parkinson’s disease

33. March 9, Mohammad Bagher Menhaj (Amirkabir University of Technology, Iran) Cognitive control (robotics) 34. February 13, Zakiyullah Romdlony (University of Groningen) stabilization with guaranteed safety of nonlinear systems 35. January 23, Riccardo Ferrari (TU Delft) Distributed fault diagnosis and cyber-attack detection in large–scale systems

83 Bernoulli Institute Annual Report

84 Bernoulli Institute Annual Report

6 Distributed Systems

Group leader: Prof. dr. A. Lazovik

Tenured staff (BI members) source fte Prof. dr. A. Lazovik RuG 1.0 Prof. dr. ir. M. Aiello RuG ?

Postdocs H. Groefsema RuG 0.5 F. Blaauw ECiDA, 0.6, GDBC 0.2

PhD students T.B Dijkhuis Hanze/RuG 0.4 (supervisor: Aiello) L. Fiorini NWO 1.0 (supervisor: Aiello) A. R. Pratama LPDP 1.0 (supervisor: Lazovik) A. Sha CSC 1.0 (supervisor: Aiello) M. Medema RuG 1.0 (supervisor: Lazovik)

Guests N. Eder, GK Software, Germany M. Muller, GK Software, Germany Prof. Y. Bai, Shanghai Polytechnic University, China L. Kai, Shanghai Polytechnic University, China D. Scherff, GK Software, Germany

85 Bernoulli Institute Annual Report

86 Bernoulli Institute Annual Report

6.1 Research Program

Distributed Information Systems are concerned with the delocalization of computation on several hosts and their coordination via message passing. Looking at today’s information systems, one notices that most of them, if not all, have some form of distribution. The key issues that emerge for research become those of addressing heterogeneity, scalability, and run-time adaptation. In the context of distributed systems, the group focuses on a number of sub-areas: Service-Oriented and Cloud Computing, Pervasive Computing and Sensor Networks. While interesting applications areas for the group are: Healthcare, Domotics and Smart Energy Systems.

6.1.1 Service-Oriented and Cloud Computing

Service-Oriented Computing (SOC) is a popular computing paradigm for building distributed information systems in which the concepts of distribution, openness, asynchronous messaging and loose coupling take a leading role. In this context, applications are built out of individual services that expose functionalities by publishing their interfaces into appropriate repositories, abstracting entirely from the underlying implementation. Published interfaces may be searched by other services or users and subsequently be invoked. The interest in SOC is a consequence of the shift from a vision of a Web based on the presentation of information to a vision of the Web as computational infrastructure, where systems and services can interact in order to fulfill users’ requests. Web Services (WS), the best-known example, are the realization of service-oriented systems based on open standards and infrastructures, extending the XML syntax. The ‘servicization’ of software envisioned with SOC has brought to the idea of Cloud Computing. In the latter approach, services are further abstracted and clustered in opaque and remote “clouds” of computational and storage services. This allows for virtually infinite scalability from the service consumer perspective, while promoting the ‘offering’ of underutilized resources on the producer’s side. Our group is active on three main lines of research: (1) Artificial Intelligence (AI) planning for taking advantage of the dynamicity of SOC and Cloud frameworks; (2) Cloud provisioning; and (3) Service-based business process management. Automatic Service Composition. Service domains constitute an application field where automated planning can significantly contribute towards achieving customizable and adaptable composition. In many cases, domains can be characterised with composite services that describe more complex situations than basic services. The composite services contain advice on how to perform efficient compositions. Among planning techniques, Hierarchical Task Network (HTN) planning supports representing such composite information and enables automatic service composition. HTN planning uses a domain description that contains methods

87 Bernoulli Institute Annual Report as possible ways of decomposing composite services. We investigated the use of HTN planning for service composition in the domains of ubiquitous computing and cloud computing. Automation and Decision Making in Buildings. The vision of a future in which environ- ments support the people occupying them requires computing facilities that could help evolve these environments. At the same time, the overall objective to save energy as much as possible should be maintained. All this can be accomplished by bringing a degree of sophistication to the processing of and reasoning over the information provided by a network of diverse devices and sensors. The processing refers to the achievement of some goal (or performing a task), and the selection and combination of tasks at run-time. In this way, the goal achievement can result in different solutions depending on the current state of the environment. Modern indoor spaces form an environment that is particularly suitable for the application of automated planning. While the environment is well structured and usually well defined, it is also partially controllable, which makes the added value gained by non-trivial automated composition and monitoring of operations a feasible and realistic task. Given such setting, automated planning can perform powerful reasoning for complex tasks which considerably advance the level of environment intelligence. We have deployed our HTN planner in the restaurant of our own office building Bernoulliborg. The planner computes a plan as a sequence of actions that control restaurant lamps in such a way that lamps are turned on only if necessary with respect to the natural light level and the presence of people in the restaurant. This enables us to achieve interesting results on energy saving. Model Checking for Business Process Variability and Compliance. Business processes are collaborations between actors which aim at achieving a specific value-added goal. When automated, a business process is modeled as interlinked sets of tasks with decision points allowing for different or parallel execution paths. Different tasks can be assigned to actors which can be either fully automated services or require human input. Business process management aims to increase the performance of companies by managing these business processes. Originally designed to support rigid and repetitive work, business processes have evolved to support loosely-coupled service compositions. In service compositions, each task is implemented as an independent, self-contained, and well-defined modular service. Although tasks in service compositions are implemented in a modular fashion, the composition itself remains rigid without any possible change. Compliance verification (i.e., confirming whether a business process is compliant with a set of rules imposed on that process) can be imposed by national law or international regulations. As a result, companies must ensure that their business processes remain compliant with regulations — regardless of process modularity or complexity — or face severe penalties and law suits. At the same time, the increasing demand for customization at both smart industry, as well as service compositions in cloud configurations, demonstrates the need for a next generation of customizable business processes. When customization in compositions is defined as ‘compliance to sets of rules describing all

88 Bernoulli Institute Annual Report possible options’ — it will enable this next generation of customizable business processes that simultaneously remain compliant with regulations.

6.1.2 Pervasive Computing and Sensor Networks.

Pervasive computing envisions a future in which computers seamlessly blend into the fabric of daily life and eventually “disappear” in the environment. Domotics and building automation consider indoor environments daily used by humans where large numbers of small, inexpensive and networked processing units are embedded into everyday objects. These units are organised and interoperate in Wireless Sensor Networks (WSNs). Applications range from support for healthy aging, for people needing medical assistance, or for saving energy by means of automatic control. WSNs Techniques for Energy Savings in Buildings. We strive to apply our smart solutions to new buildings to make them even more energy- efficient. To realise such solutions, we believe that buildings should be able to understand not only their outside and inside environment conditions (e.g., temperature, natural light, etc.) but also their occupants’ activity, thus they can adapt their operations in order to reduce their energy consumption while maintaining the same level of service. For the sake of environment monitoring and context information gathering, we rely on wireless sensor networks (WSNs) to monitor and collect context information as well as for actuating and controlling the environment accordingly. Existing buildings are responsible for more than 40% of the world’s total primary energy consumption. Office buildings are responsible for a significant fraction of the energy con- sumption and greenhouse gas emissions worldwide. Moreover, current building management systems fail to reduce unnecessary energy consumption while preserving user comfort because they are unable to cope with dynamic changes due to users’ interaction with the environment. Therefore, to cope with these dynamic changes of building’s environment and users’ activity, we study, both theoretically and practically, middleware for sensing and controlling buildings with emphasis on human activity recognition, occupancy detection, and adaptive control to reach savings of up to 35% in building energy consumption. We further explore the opportunity of utilizing mobile phones and power meters as context sources. These sources are chosen as their availability for various purposes (e.g., mobile phones to support the productivity of employees and power meters for energy monitoring in a building). Specifically, the mobile phone is used to explore room-level localization and to infer occupancy of particular rooms based on received signals, while wireless power meter nodes are used to measure electricity consumption as occupancy signs. We work on room-level

89 Bernoulli Institute Annual Report power meter, which is the measurement of aggregated consumption of all individuals in a room, to reduce sensor deployment costs as well as the level of intrusiveness. Moreover, we are studying and implement fusion approaches combining the two aforementioned modalities for improving occupancy recognition. The improved occupancy recognition (i.e., in terms of accuracy and information detail) is useful in raising energy efficiency adjusting the optimal configuration of electrical devices based on user preference and analyzing movement pattern to adjust Heating, Ventilation, and Air Conditioning (HVAC) system in a building.

6.1.3 Smart Energy Infrastructures

In 2018, the focus was on studying the effects of varying topologies by evaluating them on the test feeders based on the Monte Carlo method. From a topological point of view, we compared the performance of different topology models based on the optimization solution proposed in our previous work. We evaluated a complete graph, a random graph, a small-world graph, and a radial graph. This study provides further evidence to the potential of constructing practical and efficient distribution networks for prosumers. The result of this study was published as a conference paper on The 8th International Conference on Sustainable Energy Information Technology, May 8-11, 2018, Porto, Portugal. In the future work, it is our intention to introduce distributed energy storage systems (DESS, such as home batteries and electric vehicles) and to add the optimization of charge/discharge cycle to the model. With DESS, prosumers are enabled to decide whether the excess energy should be sold for an immediate profit or stored for later use.

6.1.4 Modern Energy Systems

To date, electricity and natural gas systems have been mostly planned and operated indepen- dently. However, today’s ambitious energy policies motivate a more integrated study of the energy system, in order to achieve energy efficiency and low-carbon footprint, while increasing the security and reliability of the system. Natural gas, indeed, has unquestionable advantages, both by an economic and environmental point of view, if compared with other fossil fuels. At the same time, increasing amount of medium and small scale renewable plants are connected to the electricity network, urging to focus on a physical and operative integration of natural gas and power sectors. A new vision of the energy system is required, where different infrastruc- tures are integrated, such as natural gas and electricity grids, heating devices and renewable (solar) small-scale plants, while taking into account end-users preferences and behavior. In 2018, we have continued a survey on the energy management problem in residential and office buildings, where both the power and the thermal demands have to be satisfied. The

90 Bernoulli Institute Annual Report work mainly focuses on the optimal scheduling of DER and loads, aiming to optimize a cost and/or emission function. We investigated the most common models for the main components of the system and optimization techniques. By surveying 69 papers, we propose guidelines and recommendations for tackling the energy management problem which could be useful for designers of energy management systems and researches investigating this field. Additionally, we developed a model for the optimal scheduling of household appliances, according to emission signals coming from the electricity distribution grid. The aim is to understand to what extent we can reduce CO2 emissions by integrating the use of multiple energy carriers to supply residential energy demands. This goal can be achieved by promoting hybrid appliances and hybrid thermal loads, with the former using multiple energy carriers alternatively and the latter being satisfied by diverse technologies. Moreover, we investigated the complexity of sustainable energy choices in daily tasks. Starting from a simple example, cooking pasta, we evaluated the environmental impact, in terms of CO2 emissions, of several alternative solutions, i.e., starting from cold or hot water, and using different (combination of) appliances. Although the emission savings are small for a single instance of an activity, simple daily choices that many people make can have a major sustainability impact. Yet, the complexity of such choices is overwhelming and our cognitive resources are too limited to cope with it. Automated systems and technologies are key elements to help and support the end-user in those decisions. Yet, those systems can achieve the desired goal only if they are accepted and used as intended by large groups of users. We propose an interdisciplinary research agenda to develop acceptable and effective automation systems to promote sustainable energy systems. As future works, we will investigate possible models to forecast CO2 signals and how to simulate the synchronization effects on a large energy system, when the majority of buildings reacts to the same CO2 signal.

6.1.5 Healthcare

In modern health care a paradigm shift is taking place towards a patient-centered approach, where patients are in control of managing their disease, often through the use of web applica- tions. For patients suffering from schizophrenia however, little has been done so far and this is often attributed to the fact that these patients have different needs with respect to the structure, content, and user interface of a web site. This is why the University Medical Center Groningen (UMCG) and the University of Groningen cooperate to design an intelligent web application specifically for this group of patients. By data mining patients’ questionnaire answers, test results, health records, and demographic information, relevant information and intelligent suggestions can be offered, personalized and localized for each patient. Another project we are working on is based on automating vector autoregression (VAR) of

91 Bernoulli Institute Annual Report patient diary data. VAR techniques allow medical professionals to see, for individual patients, for example whether depression causes a lack of activity, or whether a lack of activity causes depression. VAR is a promising technique when applied in the analysis of patient diary data, but the construction of VAR models requires expertise and time. Currently, vector autoregression on patient diary data is strictly done manually by statisticians, which causes a delay of weeks between when patients enter their data and when clinicians can interpret the results. In 2013, we have created a web application that automates most parts of the VAR approach and can provide medical professionals with quick results. This project is also in cooperation with the UMCG. A third research project is in collaboration with Healthcare group Espria, who have set up a project to include a person-centred approach for assessing the well-being of elderly people. Together with Espria we have created a mobile phone application to perform ecological momentary assessment (EMA) studies on a large group of elderly people. In EMA participants are measured for a longer period of time using questionnaires. With these questionnaires we want to provide the participants with interesting feedback about their well-being. Such feedback could provide insight into their well-being and help to sustain or even enhance it. The University Medical Center Groningen (UMCG) and the University of Groningen (RuG) are collaborating on the development of a system to conduct these questionnaires, to analyse the data, and to provide meaningful and personal feedback. The research in this project focuses on novel and effective ways to analyse the collected data and to provide feedback. In the first phase the research focused on the general population of the Netherlands (i.e., all people aged 18 and above), in a project known as HowNutsAreTheDutch? (www.hoegekis.nl). In the second phase the focus shifts towards the elderly people in the Netherlands. For this second project we are applying novel methods to analyse the data and provide feedback, but are also tailoring the questionnaires conducted to the participants, by asking only the questions relevant to the person under study. In the last phase of the project (the current phase) we also research various machine learning techniques, and determine which of them could best be used for analysis in the field of mental health. For example, as a means to determine response to treatment. For this we apply the state of the art targeted learning methodology, that aims to reconcile the gap between traditional statistical inference and modern machine learning. This part of the research is done in collaboration with the Biostatistics Departement of the University of California, Berkeley (Berkeley, USA) and MAP5 of Universite´ Descartes (Paris, France).

92 Bernoulli Institute Annual Report

6.1.6 Developmental Psychology

Early school drop out is a significant problem amongst community college students. To research and reduce early school drop out, we started the u-can-act project. u-can-act is a research projects in which we investigate the effective methods to prevent early school drop out in community college students. In this project we collaborate with three Dutch student supervisory agencies: Het Buro (Leeuwarden), Mijn School (Doetinchem), and the Plusgroep (The Hague). The project itself revolves around three main research-questions: what are thecrucial developments with which young adolescents make the decision to drop out of school?, how can supervisory agencies most effectively intervene on these developments in order to prevent early school drop out?, and how can this knowledge be optimally applied in policy making? In u-can-act we answer these questions using a novel research methodology, by applying an Ecological Momentary Assessment (EMA) study. In this EMA study (which takes approxi- mately six months) we ask a group of supervised students to complete a weekly questionnaire provided to them via a Web application. Besides measuring the students, we also ask their supervisor to complete a series of weekly questionnaires about the students they supervise (in the same period). This way we measure both the students’ development over time, as well as the actions and interventions the supervisors performed to steer this development. Such measurements provide a novel insight into the meso level dynamics of the student-supervisor interactions.

6.1.7 ECiDA

Modern data analysis platforms all too often rely on the fact that the application and underlying data flow are static. That is, such platforms generally do not implement the capabilities to update individual components of running pipelines without restarting the pipeline, and they rely on data sources to remain unchanged while they are being used. However, in reality these assumptions do not hold: data scientists come up with new methods to analyze data all the time, and data sources are almost by definition dynamic. Companies performing data science analyses either need to accept the fact that their pipeline goes down during an update, or they should run a duplicate setup of their often costly infrastructure that continues the pipeline operations. In this research we present the Evolutionary Changes in Data Analysis (ECiDA) platform, with which we show how evolution and data science can go hand in hand. ECiDA aims to bridge the gap that is present between engineers that build large scale computation platforms on the one hand, and data scientists that perform analyses on large quantities of data on the other, while

93 Bernoulli Institute Annual Report making change a first-class citizen. ECiDA allows data scientists to build their data science pipelines on scalable infrastructures, and make changes to them while they remain up and running. Such changes can range from parameter changes in individual pipeline components to general changes in network topology. Changes may also be initiated by an ECiDA pipeline itself as part of a diagnostic response: for instance, it may dynamically replace a data source that has become unavailable with one that is available. To make sure the platform remains in a consistent state while performing these updates, ECiDA uses a set of automatic formal verification methods, such as constraint programming and AI planning, to transparently check the validity of updates and prevent undesired behavior.

6.1.8 Automated Analysis of Human Perfomance Data

Furthermore, we also contributed to to developing automated data analysis systems for small and huge amounts of data and create a set of tools which support automated analysis, algorithm construction,and intervention in reality. The overall research question is:“How can one develop a validated automated data analysis system to explore of quantitative health data, in relation with (general) contextual data and characteristics of the individual, that reveals patterns, features and anomalies which assists the expert or system in his reasoning and intervention possibilities”. The contribution to the science is twofold. First: this research contributes to the field of methodology on Knowledge Discovery in Databases (KDD) and builds theory on how to construct a KDD system for huge amounts of data. Second: the set of tools is a system which saves time on data preparation, pattern recognition, and statistical inference on sensor data in combination with contextual data and contributes to science in a practical way, as experts can focus on performing research, rather than spend time on tedious data preparation, pattern recognition and statistical inference. Besides the scientific contributions, the research also makes several noteworthy societal contributions. Firstly, by automating the cumbersome and time consuming part of research on quantitative data enables to find information faster and gain knowledge to intervene in reality, to improve the human performance. Secondly, the development of the system aids the identification of requirements and possibilities for the development of valid systems for (research) purposes on data. In 2018 a paper on combining algorithms for the prediction on daily steps of employees was presented at the eTELEMED 2018 conference. We published a manuscript on applying machine learning to personalized physical activity coaching to the peer reviewed journal Sensors. A conference paper on the influence of workingdays vs leisure time on trained algorithms was submitted and accepted for the eTELEMED 2019 conference.

94 Bernoulli Institute Annual Report

In the future we are working on the prediction of running injuries from training load. Another project is the implementation of the online super learner concept to combine the scientifical statisctical field with the machine learning approach within the health domain. A third project is about the realisation and testing of an automated virtual coaching system on physical activity.

6.1.9 Large-Scale Combinatorial Optimisation

Finding the optimal solution to solve a given problem is an important objective that arises within many different domains, ranging from making the most optimal use of resources for a set of computational tasks to transporting packages in the least amount of time with the lowest possible cost. Constraint satisfaction problems offer a powerful mathematical framework to define such optimisation problems without the need to specify explicitly how the problem must be optimised. Unfortunately, these problems are NP-complete, meaning that, in the worst case, an algorithm must consider all possible combinations of potential solutions. This computational complexity makes it, in general, impossible to apply to large-scale domains. However, sometimes problems have a particular structure that can be used to find a solution far more efficiently than the worst-case exponential complexity. This research focuses on a class of problems with such a structure, which is a natural model for certain large-scale domains such as smart building systems, and an algorithm that can solve problems of this class efficiently. After the principles of the algorithm have been established and the initial performance has been tested on synthetic problems, the aim is to apply the optimisation techniques on real-world problems to verify the results. Developing an algorithm that is able to determine whether a given constraint satisfaction problem belongs to the previously mentioned class of problems is another useful direction of research, as this makes it possible to make a more informed decision about which algorithm is most suitable to solve a given problem.

6.2 Research subjects

M. Aiello: Service-Oriented Computing, Ubiquitous Computing, Smart Energy Systems. F.J. Blaauw: Data science, big data, distributed systems, machine learning, ecological mo- mentary assessments T.B. Dijkhuis: Machine learning, statistics, health, human performance L. Fiorini: energy management systems, scheduling, emissions, renewable energy, hybrid loads H. Groefsema: Business Process Management, Service Composition, Variability Manage- ment, Compliance Verification, Model Checking. A. Lazovik: Automated Service Composition, Monitoring and Repair, Automated Planning,

95 Bernoulli Institute Annual Report

Cloud Computing. A.R. Pratama: Context-awareness, Occupancy recognition, Non-Intrusive Load Monitoring. B. Setz: Internet of Things, Sustainable Data Centers, Cloud Computing, Green Computing, Machine Learning. A. Sha: Smart Grid, Topology, Energy Distribution, Distributed Energy Generation, Monte Carlo Simulation. M. Medema: Discrete Optimisation, Constraint Satisfaction Problems, Combinatorial Opti- misation.

6.3 Publications

Books

M. Aiello, The Web Was Done by Amateurs: A Reflection on One of the Largest • Collective Systems Ever Engineered, Springer, 2018.

Dissertations

Blaauw, F. J. The non-existent average individual: Automated personalization in • psychopathology research by leveraging the capabilities of data science, Promotor: M.Aiello, Faculty of Science and Engineering, University of Groningen, 12 February 2018, 294 pages.

Articles in scientific journals

Bos, F. M., Blaauw, F. J., Snippe, E., Van der Krieke, L., de Jonge, P., and Wichers, M. • C. Exploring the emotional dynamics of subclinically depressed individuals with and without anhedonia: an experience sampling study. Journal of Affective Disorders, 228, 2018, 186 – 193.

H. Groefsema, N. R. T. P. van Beest and M. Aiello, A Formal Model for Compliance • Verification of Service Compositions, IEEE Transactions on Services Computing, 11 (3), 2018, 466 - 479.

F. M. Bos, F. J. Blaauw, E. Snippe, L. van der Krieke, P. de Jonge and M. Wichers, • Exploring the emotional dynamics of subclinically depressed individuals with and without anhedonia: An experience sampling study, Journal of Affective Disorders, (228), 2018, 186 - 193.

96 Bernoulli Institute Annual Report

L. Fiorini, M. Aiello, D. Poli and P. Pelacchi, Topological Considerations on the Use of • Batteries to Enhance the Reliability of HV-Grids, Journal of Energy Storage, 2018, 316 - 326. A. R. Pratama, W. Widyawan, A. Lazovik and M. Aiello, Multi-User Low Intrusive • Occupancy Detection, Sensors, 18 (3), 2018, 796. L. Fiorini and M. Aiello, Household CO2-efficient energy management, Energy Infor- • matics, 2018. T. B. Dijkhuis, F. J. Blaauw, M. W. van Ittersum, H. Velthuijsen and M. Aiello , • Personalized Physical Activity Coaching: A Machine Learning Approach, Sensors, 18 (623), 2018.

Refereed articles in conference and workshop proceedings

Blaauw, F., Emerencia, A. C., den Hartigh, J. R., Milovanovic,´ M., Stoter, I., and de • Jonge, P. Predictions from the cloud: using data science to predict sports performance, Abstract from Science and Engineering Conference on Sports Innovation, Groningen, Netherlands, 2018. A. Sha and M. Aiello, Topological Considerations on Decentralised Energy Exchange • in the Smart Grid, The 9th International Conference on Ambient Systems, 2018. A. R. Pratama, F. J. Simanjuntak, A. Lazovik and M. Aiello, Low-power Appliance • Recognition using Recurrent Neural Networks, Applications of Intelligent Systems, (310), 2018, 239 - 250. M. Bessani, R. R. M. Ribeiro, G. A. Pagani, M. Aiello and C. D. Maciel, Robustness • of reconfigurable complex systems by a multi-agent simulation: Application on power distribution systems, Annual IEEE International Systems Conference, 2018, 1 - 6.

M. Kalksma, B. Setz, A. R. Pratama, I. Georgievski and M. Aiello, Mining Sequential • Patterns for Appliance Usage Prediction, International Conference on Smart Cities and Green ICT Systems, 2018

97 Bernoulli Institute Annual Report

6.4 External funding and collaborations

External funding: active large projects.

Acronym Name Funding Programme PI Agency BIGS Beijing Groningen Smart en- NWO JSTP Smart Aiello ergy cities Energy in Smart Cities ECiDA Evolutionary changes in Dis- NWO Commit2Data Lazovik tributed Analysis MERGE Integration of Gas and Elec- NWO ESI-Pose Aiello tricity Nerdalize Heating Houses with Com- STW STW Take-off Lazovik puting Power in the Form of Grant Radiators NextGen Next Generation Smart Data NWO Indo Dutch Sci- Aiello SmartDC Centers ence Industry Collaboration in Computer Science OurEnergy OurEnergy TKI Urban Energy Lazovik u-can-act Voortijdig schoolverlaten NRO Beleidsgericht P. de voorkomen: een focus op het Onderwijsonder- Jonge proces van schoolverlaten en zoek een methodiek om dit proces tijdig bij te sturen.

The DS group has active a number of research projects funded mostly by NWO in areas tied to smart energy systems, sustainability, collaboration with industry (see Table External funding). A number of PhD positions are also funded by foreign funding agencies (see Table PhD study Grants).

External funding: PhD study grants.

98 Bernoulli Institute Annual Report

Acronym Name Funding Agency Recipient PI CSC CSC-RUG joint China Scholarship Sha Aiello scholarships pro- Council gramme LPDP Indonesia Endow- Ministry of Finance Pratama Aiello ment Fund for of the Republic of In- Education donesia

External Collaborations

Most notable collaborations were active with the following individuals and institutions: Prof. P. Pelacchi and Prof. M. Poli, University of Pisa, Jian Yang, Michael Zheng, Macquarie University, Sydney, Australia, Robin Hagemans, Pieter den Hamer, and Floran Stuijt, Alliander N.V., Tarun Kumar, Aanchal Aggarwal, Mark Lavin, and Abhishek Raman all, IBM Research, Marti Rosas-Casals, Polytechnic University of Catalonia, Lingen Luo, Shanghai Jiao Tong University, University of California, Berkeley (UCB), Prof. J. Slaets, Espria Academy, Prof. P. de Jonge, UMCG, Ester Kuiper and Chantal Bosman, Lifely, Dr. Blagoj Ristevski, St. Kliment Ohridski University in Bitola, V. Dinesh Reddy and Prof. G R Gangadharan, the Institute for Development and Research in Banking Technology (IDRBT), Hyderabad, India, Collaboration with Dr. VRK Rao, Subrahmanya, Assistent Vice President Technology at Cognizant Technology Solutions. A. Berfu Unal and Ellen Van Der Werff, Environmental Psychology RUG, Giovanni Squillero, Politecnico di Torino, Alberto Tonda, French National Institute for Agricultural Research, Giovanni Iacca, INCAS3, Dr. Guram Beshanshvili, New Mexico State University, Prof. Dr. Dustdar, Technical University of Vienna, Prof. Dr. Wim van Gemert, Hanzehogeschool, Prof. Dr. Gottlob, Technical University of Vienna, Prof. Dr. George Huitema, RUG and TNO, Peter Kamphuis, Hanzehogeschool, Groningen, Prof. Dr. Eiter, Technical University of Vienna, Prof. Dr. Han Slootweg, TU/e and Enexis, Prof. Dr. Stefan Tai, Karlsruhe Institute of Technology, Prof. Dr. Apt, Arbab CWI, Prof. Dr. Michael Beigl, KIT, Prof. Dr. Daniele Nardi, UoR, Prof. Dr. Bernhard Nebel, FU, Prof. Dr. Roberto Baldoni, UoR, Dr. Massimo Mecella, UoR, Prof. Dr. HT Yang, National Cheng Kung University, Dr. Rix Groenboem, Parasoft, Hemmo Halzebos, Enexis, Dr. Heiko Ludwig, IBM, Wico Mulder, Logica, H. Zwaal, TenneT, Emiliano Binotti, Fluid Solutions, Dr. Paul Shrubsole, Philips Research Laboratories, Dr. Chun Yu Chen, ITRI, Jose J de las Heras, Advantic Sistemas Y Servicios S.L., Prof Dr. Marti Rosas-Casals, Polytechnic University of Catalonia, Will West Control4 Corporation, Prof. Dr. Christian Claudel, King Abdullah University of Science and Technology, Prof. Dr. Kenji Tei, Japanese National Institute of Informatics. Prof. M. van der Laan, University of California, Berkeley (UCB), Prof. A. Chambaz, Universite´ Descartes, Prof. P. de Jonge, University of Groningen. V. Dinesh

99 Bernoulli Institute Annual Report

Reddy and Prof. G R Gangadharan, the Institute for Development and Research in Banking Technology (IDRBT), Hyderabad, India, Collaboration with Dr. VRK Rao, Subrahmanya, Assistent Vice President Technology at Cognizant Technology Solutions. Prof. P. Pelacchi and Prof. D. Poli, University of Pisa.

100 Bernoulli Institute Annual Report

7 Fundamental Computing

Group leader: Prof.dr. G.R. Renardel de Lavalette

Tenured staff (BI members) source fte Prof.dr. G.R. Renardel de Lavalette RUG 1.0

Tenure Track Dr. J.A. Perez´ Parra RUG 1.0

Tenured staff (other) Dr. A. Meijster RUG (USSE) 1.0

Emeritus Prof.dr. W.H. Hesselink

PhD students A. Arslanagic bursary (supervisor: J.A. Perez´ Parra and G.R. Renardel de Lavalette)

M. Cano RUG 1.0 (supervisor: J.A. Perez´ Parra and G.R. Renardel de Lavalette)

J. Paulus NWO 1.0 (supervisor: J.A. Perez´ Parra and G.R. Renardel de Lavalette)

Guests I. Castellani, Imperial College London, UK J. Dedeic,´ University of Novi Sad, Serbia C. Di Giusto, University of Nice, France E. Buitrago D. Nantes, University of Brasilia, Brazil

101 Bernoulli Institute Annual Report

102 Bernoulli Institute Annual Report

7.1 Research Program

The objective of this programme is to contribute to the understanding of the logical and mathematical foundations of computing science and to realize a two-way transfer between this fundamental research and more applied subdisciplines of computing science. Our research focuses on formal methods, which are based on concepts and theories from discrete mathemat- ics and logic. They are applied to enhance the reliability of computer systems and computer software, and also to further the understanding of the possibilities of computing in general. The following themes are studied: formal methods for concurrent systems, programming methodology, multi-agent systems, mathematical logic. In formal methods for concurrent systems, the main issues are specification and analysis. The research for specification focuses on process calculi, formal languages that express the interaction of concurrent processes in a compositional style, while the research for analysis focuses on type systems for process calculi, often referred to as behavioral type systems. Behavioral types codify the protocols that channels in concurrent processes should implement. Coupling process calculi with behavioral types gives rise to a compositional approach to verify safety and liveness properties for concurrent, message-passing systems. A well-studied class of behavioral types is session types; this way, session-typed concurrency refers to a model in which concurrent processes follow protocols specified as session types. In mathematical logic, we focus on the proof theory of equational logic and Horn logic, and on intuitionistic logic. Intuitionism was created by L.E.J. Brouwer; its logic, formalized by A. Heyting, reappeared as the foundation of type theory with applications in programming and theorem proving. Equational logic is the formalization of algebraic equational reasoning, used in tools like Mathematica. Horn logic is the logic of formulae of the form A1 ... A B; ∧ ∧ n → it is the basis of the logic programming language Prolog. We focus on fundamental properties like completeness (is a given proof system strong enough to prove all true statements?) and exactness (does a certain model correspond to the structure of a logic?).

7.2 Research subjects

Cano: session-typed concurrency, behavioural types, sychronous programming. Hesselink: design and verification of concurrent and geometric algorithms. Paulus: concurrency theory, typed process calculi, behavioural type systems. Perez´ : concurrency theory, semantics, process calculi, type systems. Renardel: proof theory of Horn logic; intuitionistic logic.

103 Bernoulli Institute Annual Report

7.3 Publications

Articles in scientific journals

A. Aravind, W.H. Hesselink, Group mutual exclusion by fetch-and-increment, ACM • Transactions on Parallel Computing 12 (2018)

P.A. Buhr, D. Dice, W.H. Hesselink, High-contention mutual exclusion by elevator • algorithms, Concurrency and Computation: Practice and Experience 30, issue 18 (2018)

A. Francalanza, J. Perez,´ and C. Sanchez,´ Runtime Verification for Decentralised and • Distributed Systems, Springer, 2 (2018) 176–210

J. van de Gronde, W.H. Hesselink, Conditionally complete sponges: new results on • generalized lattices, Indagationes Mathematicae 11 (2018)

W.H. Hesselink, Tournaments for mutual exclusion: verification and concurrent com- • plexity, Formal Aspects of Computing 29 (2017) 833–852

W.H. Hesselink, The quartet spaces of G. ’t Hooft, Indagationes Mathematicae 29 (2) • (2018) 628–632

W.H. Hesselink, P.A. Buhr, D. Dice, Fast mutual exclusion by the triangle algorithm, • Concurrency and Computation: Practice and Experience 30, issue 4 (2018)

Gerard R. Renardel de Lavalette, Interpolation in propositional Horn logic, Journal of • Logic and Computation 28 (2018) 1189–1215

Articles in conference proceedings

D. Nantes, J. Perez,´ Relating process languages for security and communication correct- • ness (extended abstract), 38th IFIP WG 6.1 International Conference FORTE (2018) 79–100 F. de Vries, J.A. Perez,´ Reversible Session-Based Concurrency in Haskell, in: M. Palka, M. Myreen (eds.), Trends in Functional Programming (TFP 2018), Lecture Notes in Computer Science, vol 11457, Springer-Verlag, https://doi.org/10.1007/978-3-030- 18506-0 2

104 Bernoulli Institute Annual Report

7.4 External funding, collaboration and internationalization

Hesselink collaborates with P.A. Buhr (University of Waterloo, Canada) and D. Dice (Oracle Labs, Burlington, USA) on design and verification of mutual exclusion algorithms for shared memory systems, and with Alex Aravind (University of Northern British Columbia) on algo- rithms for group mutual exclusion. In 2018, Perez´ served as member of the management committees of the EU COST Actions IC1402 (ARVI: Runtime Verification Beyond Monitoring) and IC1405 (Reversible Computa- tion - Extending Horizons of Computing). Perez´ is Co-PI in Project SuCCeSS: Securit´ e,´ adaptabilite,´ et information temporise´ dans les Communication-Centric Software Systems (Security, adaptability, and time in Communication- Centric Software Systems), funded by CNRS (PICS program) to support exchanges between Groningen and France. Perez´ collaborates on diverse aspects of concurrency theory and behavioral types with: L. Caires (University NOVA of Lisbon, Portugal), I. Castellani (INRIA, France), A. Ciabattoni (TU Wien, Austria), M. Dezani-Ciancaglini (University of Turin, Italy), C. Di Giusto (Univer- sity of Nice, France), O. Dardha (University of Glasgow, UK), D. Kouzapas (University of Cyprus, Cyprus), C. Mezzina (IMT Lucca, Italy), D. Nantes (University of Brasilia, Brazil), J. Pantovic (University of Novi Sad, Serbia), C. Rueda (Universidad Javeriana, Colombia), N. Yoshida (Imperial College London, UK)

7.5 Further information

Hesselink is a member of the Editorial Board of the international scientific journal Science of Computer Programming.

105 Bernoulli Institute Annual Report

106 Bernoulli Institute Annual Report

8 Information Systems

Group leader: Prof. dr. Dimka Karastoyanova Tenured staff (BI members) source fte

Prof. dr. D. Karastoyanova RUG 1.0

Tenure track assistant professors Dr. G. Azzopardi RUG 0.8

PhD students X. Wang (since 01-07-2018) CSC 1.0 (supervisor: Azzopardi, Karastoyanova) L. Song (since 01-07-2018) RUG 1.0 (supervisor: Azzopardi, Karastoyanova) A. Alsahaf (since 01-02-2016) RUG 1.0 (supervisor: Azzopardi, Petkov) A. Shi RUG 1.0 (supervisor: Azzopardi, Petkov)

Guests S. Saikia, University of Leon, Leon, Spain Dr. J. Ellul, University of Malta, Malta Dr. O. Falzon, University of Malta, Malta

107 Bernoulli Institute Annual Report

108 Bernoulli Institute Annual Report

8.1 Research Program

Our mission is to significantly advance the state-of-the-art in Information Systems through focused fundamental and applied research in Adaptive Information Systems, Data-driven Information Systems and Information Assurance and Security, while explicitly considering their interconnections and applicability in real-world scenarios. We are contributing to both fundamental and applied research and are excited about interdisciplinary topics involving industry and academic organisations. In particular, our research revolves around the following topics:

Adaptive Information Systems with focus on service-oriented process automation and flexi- ble workflow management as the backbone of modern enterprise systems. Special attention is paid on both theoretical aspects of business processes and technological aspects of the corresponding middleware and their software architecture and realizations. In this research area we currently focus on the topics: a) adaptive service orchestrations and choreographies, b) process performance monitoring and improvement, monitoring of KPIs of both orchestrations and choreographies, c) Data-driven process adaptation, IoT and Data Analytics and d) Blockchain enabled adaptive service choreographies for business and scientific applications

The Data-driven information systems area focuses on image processing, pattern recognition and machine learning in the context of the Internet of Things (IoT). In particular, the research interests are in the fields of:

Medical image analysis - computer aided diagnosis systems • Person identification by means of vision-based biometric features • Smart farming • Predictive analysis in sports science • Brain-inspired computer vision • Autonomous health monitoring systems with IoT • Blockchain and AI •

Information Assurance and Security targets the design, analysis, formal/semi-formal tech- niques for verification and implementation of security systems tailored to access control, key management and computation over encrypted data. Information systems often deal with (collect, store and process) sensitive data. Our group also performs research on data-centric se- curity and privacy (S&P) at the intersection of authentication and authorization infrastructures,

109 Bernoulli Institute Annual Report privacy-enhancing technologies (PET) and trust management. With respect to applications, our initial plan is to focus on addressing societal and economical challenges (with respect to S&P) in the area of Groningen by collaborating with the local partners and expanding the developed concepts to national/international levels. The particular research areas include:

Formal Methods for (S&P) Analysis • Key Management and computation over encrypted data • PETs applied to (collaborative) machine learning • S&P for Medical Informatics (e.g. genomic privacy) • Cloud, IoT and Edge Security • Applications of Blockchain (with a focus on S&P) to various domains •

8.2 Overview of scientific results

Flexible Choreographies and Provenance of Scientific Workflows BPM research opened numerous opportunities for synergies with blockchains in different domains and in particular in the scope of collaborative processes. In the eScience domain however there is a need to support a different type of collaboration where adaptation is essential part of that collaboration. Scientists demand support for trial-and-error experiment modeling in collaboration with other scientists and at the same time, they require reproducible experiments and results. The first aspect has already been addressed using adaptable scientific choreographies, as our previous research shows. We therefore worked towards identifying and evaluating potential approaches, concepts and architectures for combining adaptable scientific choreographies with blockchain platforms for the purposes of enabling trust among collaborating scientists. In future we will focus on creating a system that enables provenance of scientific workflows in a generic manner and allows the use of different types of storage solutions and identifying the research questions driving the future research in security models for flexible workflows in the Cloud.

Enhancing Business Process Flexibility by Flexible Batch Processing In the scope of a research collaboration we identified the need for a more flexible processing of batched work in processes and designed a solution that both enhances the flexibility of processes in general and improves the flexibility of batch processing by workflow management

110 Bernoulli Institute Annual Report systems in particular. Our approach provides an architecture and design of a workflow management system that supports three strategies for enabling flexible batch processing.

Computer-aided diagnosis systems In collaboration with the University of Malta and the Indian Institute of Technology Hyderabad we proposed techniques to build computer-aided diagnostic systems for diabetic retinopathy and retinopathy of prematurity, which has a high prevalence in pre-term infants. For the former system we published an article in the proceedings of CAIP2019 and for the latter we have an article under review.

Nature-inspired systems The visual system of the brain has always inspired scientists to build computational models that can push forward the field of computer vision. In collaboration with the Intelligent Systems group we developed a trainable operator for the detection of curvilinear structures in images. It includes the concept of push-pull inhibition, a phenomenon that is exhibited in many neurons in the brain. This type of inhibition makes the operator very robust to different types of noise. The work is now published on IEEE Transactions on Image Processing. Moreover, together with the PhD student Liang Song, we are investigating a new family of deep learning networks that are inspired by the concepts of sparsity and neuro-evolution nature-inspired processes. We expect these networks to be configurable by non-differentiable operators and as a result to be more robust to adversarial attacks.

8.3 Research subjects

G. Azzopardi: (Brain-inspired) Pattern Recognition, Predictive Analysis. D. Karastoyanova: Service-oriented architectures, middleware, workflow management and BPM, adaptive systems, provenance of flexible scientific workflows. L. Song: Brain-inspired pattern recognition. X. Wang: Autonomous health monitoring with pattern recognition.

8.4 Publications

Dissertations

Andreas Weiss. Flexible Modeling and Execution of Choreographies” September 28th, •

111 Bernoulli Institute Annual Report

2018, University of Stuttgart. Promotor: D. Karastoyanova.

Edited Books

J. Miranda Dias, A. Bandera, G. Azzopardi, R. Marfil, “Cooperative and Social Robots: • Understanding Human Activities and Intentions (COBOT- UHAI)”, Pattern Recognition Letters, 2018, doi: 10.1016/j.patrec.2018.07.007

Articles in refereed journals

1. G. Azzopardi, A. Greco, A. Saggese, M. Vento, “Fusion of domain-specific and trainable features for gender recognition from face images”, IEEE Access, vol. 6(1), pp.24171-24183, 2018.

2. A. Alsahaf, G. Azzopardi, B. Ducro, E. Hanenberg, R. Veerkamp, N. Petkov, “Predic- tion of slaughter age in pigs and assessment of the predictive value of phenotypic and genetic information using random forest”, Journal of Animal Science, in print, 2018

3. C. Shi, G. Azzopardi, D. Zillikens, E. Schmidt, G.F.H. Diercksr, J. Guo, J.M. Meijer, M. Jonkman, N. Petkov, “Detection of u-serrated patterns in direct immunofluorescence images of autoimmune bullous diseases by inhibition-augmented COSFIRE filters”, International Journal of Medical Informatics, 2018.

Articles in refereed conference proceedings

1. A. Bonnici, D. Bugeja, G. Azzopardi, “Vectorisation of sketches with shadows and shading using COSFIRE filters”, DocEng, Halifax, 2018 – Nominated for the best paper award

2. A. Bonnici, J. Abela, N. Zammit, G. Azzopardi, “Localisation, Recognition and Ex- pression of Ornaments in Music Scores”, DocEng, Halifax, 2018

3. G. Azzopardi, P. Foggia, A. Greco, A. Saggese, M. Vento, “Gender recognition from face images using trainable shape and colour features”, ICPR, Beijing, 2018

4. F. Abadi, J. Ellul, G. Azzopardi, “The Blockchain of Things, Beyond Bitcoin: A Systematic Review”, The 1st International Workshop on Blockchain for the Internet of Things 2018 – 2018 IEEE Blockchain – BIoT, 2018

112 Bernoulli Institute Annual Report

5. A. Alsahaf, G. Azzopardi, B. Ducro, R.F. Veerkamp and N. Petkov, “Predicting Slaugh- ter Weight in Pigs with Regression Tree Ensembles.”, Applications of Intelligent Systems – Proceedings of the 1st International APPIS Conference 2018, Frontiers in Artificial Intelligence and Applications 310 (2018), 1-9 . IOS Press, Amsterdam.

6. A. Apap, L. Fernandez-Robles´ and G. Azzopardi, “Person Identification with Reti- nal Fundus Biometric Analysis Using COSFIRE Filters”, Applications of Intelligent Systems – Proceedings of the 1st International APPIS Conference 2018, Frontiers in Artificial Intelligence and Applications 310 (2018), 10-18 . IOS Press, Amsterdam.

7. J. Buhagiar, N. Strisciuglio, N. Petkov and G. Azzopardi, “Automatic Segmentation of Indoor and Outdoor Scenes from Visual Lifelogging.”, Applications of Intelligent Systems – Proceedings of the 1st International APPIS Conference 2018, Frontiers in Artificial Intelligence and Applications 310 (2018), 194-202 . IOS Press, Amsterdam.

8. N. Strisciuglio, G. Azzopardi, N. Petkov, “Robust curvilinear detection operator”, ECCVW Proceedings, in print, 2018

9. M. Spiteri, G. Azzopardi, “Customer churn prediction for a motor insurance company”, 6th IWDS, Berlin, ICDIM Proceedings, in print, 2018

10. A. Alsahaf, G. Azzopardi, B. Ducro, R. F. Veerkamp, N. Petkov, “Assigning pigs to uniform target weight groups using machine learning”, World Congress on Genetics Applied to Livestock Production (WCGALP), Auckland, New Zealand, 2018

11. L. Pufahl, D. Karastoyanova. Enhancing Business Process Flexibility by Flexible Batch Processing. In Proceedings of CoopIS 2018.

Abstracts in Symposia

1. A. Alsahaf, G. Azzopardi, Nicolai Petkov, “Estimation of live muscle scores of pigs with RGB-D images and machine learning”, FAIR 2018, Wageningen

2. A. Bhole, M. Biehl, G. Azzopardi, “Automatic recognition of Holstein cattle using non-invasive computer vision approach”, FAIR 2018, Wageningen

3. A. Neocleous, G. Azzopardi, M. W. Dee, Signal Processing for the Identification of Miyake Events, 23rd International Radiocarbon Conference, Trondheim, Norway, 2018

4. A. Neocleous, G. Azzopardi, M. W. Dee, Identification of Possible Miyake Events using COSFIRE Filters, 23rd International Radiocarbon Conference, Trondheim, Norway, 2018

113 Bernoulli Institute Annual Report

5. Constantinos Loizou, Dimka Karastoyanova, Christos N. Schizas: Measuring the Impact of Blockchain on Healthcare Applications. In Proceedings of APPIS 2019 (Applications of Intelligent Systems), January 2019, 7-9 January 2019 in Las Palmas de Gran Canaria, Spain

Other publications

D. Karastoyanova,& Stage, L. (2018). Towards Collaborative and Reproducible Scien- • tific Experiments on Blockchain. 144-149. In Proceedings of BIOC’18 @ CAiSE 2018, Tallinn, Estonia. DOI: 10.1007/978-3-319-92898-2-12

8.5 External funding and collaboration

External funding

4NSEEK: Azzopardi is member of an international consortium that received EU funding • (1.3M) for the project 4NSEEK “Forensic Against Sexual Exploitation of Children” through the financing program Internal Security Fund - Police, ISFP-2017-AG-CYBER call. George received EUR73k which he used to establish a PhD sandwich position in collaboration with the University of Leon. The new PhD candidate, Guru Swaroop, will start on 1 August 2019.

Autonomous Health Monitoring with Pattern Recognision. Funding: CSC for the PhD • candidate Xueyi Wang - PhD position from 2018 to 2022. Xueyi is under the supervision of George Azzopardi and Dimka Karastoyanova.

D. Karastoyanova was granted the NWO Westerdijk Fellowship funding. •

External collaboration

G. Azzopardi is involved in the following ongoing collaborations:

Azzopardi was an invited lecturer for the 3-day course Predictive Analysis in the Data • Science program by the IT Academy of North Netherlands (ITANN)

Intelligent Systems Group of FSE - he shares two PhD students with Prof. Nicolai • Petkov; Ahmad Alsahaf and Astone Shi

114 Bernoulli Institute Annual Report

Isotope Research Group of FSE - Research on data science approaches for radiocarbon • dating

Ophthalmology (UMCG) - Computer-aided diagnosis system for Glaucoma detection - • Prof. Nomdo Jansonius

Dermatology (UMCG) - Computer-aided diagnosis system for the detection of skin • bullous diseases - Prof. M. Jonkman

University of Wageningen, CRV, TopPigs and Hendrix Genetics - they collaborate on • the SmartBreed project which was funded by the Breed4Food of STW.

University of Salerno, Italy - Face Analysis - Prof. Mario Vento • University of Leon, Spain and University of Malta - EU 4NSEEK Project • University of Malta, Blockchain and Internet of Things - Dr. Joshua Ellul •

D. Karastoyanova is involved in the following collaborations with:

Hasso Plattner Institut, University of Potsdam - Concepts and Realizations of Flexible • Workflows and Batch Processing - Dr. Luise Pufahl

Kuhne¨ Logistics University (The KLU), Hamburg - KPIs in Humanitarian Logistics - • Prof. Maria Besiou

TNO - Blockchain as a technological enabled of the Industrial Data Space - Dr. Harrie • Bastiaansen and Simon Dalmolen.

Kuehne & Nagel AG - Predictive Analytics for Airfreight in PharmaLogistics - Jon • Chapman (Vice President Pharma Healthcare, Switzerland) and Felix Jacubasch (Senior Manager Business Intelligence & Analytics, Hamburg, Germany)

Karastoyanova was an invited lecturer for the 3-day course Data Analytics in the Data • Science program by the IT Academy of North Netherlands (ITANN).

In collaboration with Prof. M. Besiou and the NGO Plan International, D. Karastoy- • anova initiated a research project on the topic ”Modelling of KPIs for Processes in Humanitarian Logistics”. In the scope of this research a MSc Thesis has been supervised in collaboration.

Further information

115 Bernoulli Institute Annual Report

Azzopardi is associate editor of the Elsevier journals Pattern Recognition and Pattern • Recognition Letters. Morover, he served on the program committee of several con- ferences, notably NIPS, ICML, CAISE, DocEng, APPIS, and VISAPP. Similarly, he served as a reviewer to the following journals: IEEE Transactions on Image Processing, IEEE Access, and Medical Image Analysis, among others. Karastoyanova has served on the Program Committees of CAiSE 2018 Forum, the • Symposium SummerSoC2018, the Business Process Management (BPM) conference 2018, the EDOC 2018 conference, the ICSoC 2018 conference and the research Track on Microservices, DevOps and Service Oriented Architecture (MiDOS1) at SAC 2019. She was serving as a reviewer for the following journals: Concurrency and Computation- Practice and Experience, International Journal of Simulation and Process Modelling (IJSPM), Journal of Information Systems, Journal of Humanitarian Logistics and SCM (JHLSCM), ACM Transactions on Software Engineering and Methodology (TOSEM). She also participates in the GEBC (Groningen Engineering Business Centre) D. Karas- toyanova was on the reading and defense committee of the PhD of Ekaterina Bazhenova, HPI, University of Potsdam, 16.04.2018. In collaboration with Dr. Luise Pufahl, HPI, University of Potsdam, D. Karastoyanova • organised the 1st Workshop on Flexible Advanced Information Systems (FAiSE) at the CAiSE 2018 conference on Advanced Information Systems Engineering, Tallin, Estonia - https://bpt.hpi.uni-potsdam.de/FAiSE18

Invited Talks

D. Karastoyanova was invited to present during the German Logistics Congress (BVL • Kongress), Berlin, 16-17 Oct 2018 at the LOG.Camp part of the Congress at the Kick-off of the BVL-Hackathon. Topic: ”Data-driven Decisions for Smart Enterprises” D. Karastoyanova gave a presentation at the Annual Meeting of GEC and GEBS • (Groningen Engineering Center and the Groningen Engineering Business Center). Topic: “Flexible Processes for Business and eScience Applications” in the session on Software Intensive Software. 29.01.2019, Groningen

1https://midos2019.sdu.dk/

116 Bernoulli Institute Annual Report

9 Intelligent Systems

Group leader: Prof.dr.sc.techn. N. Petkov

Tenured staff (BI members) source fte Prof. Dr. M. Biehl RuG 1.0 Prof. Dr.sc.techn. N. Petkov RuG 1.0 Dr. M.H.F. Wilkinson RuG 1.0

Tenure track staff (BI members) source fte Dr. K. Bunte (Rosalind Franklin Fellow) RUG 1.0

Postdocs source fte Dr. N. Strisciuglio EU 1.0

Long term visitors source fte Dr. Li Guo PR China 0.0

PhD students A. Alsahaf NWO-STW 1.0 (supervisors: Petkov, Azzopardi) M. Babai CIT, RUG external (supervisors: Wilkinson, Petkov) D. Fernandez´ Chaves Univ. Malaga, RUG UE 1.0 (supervisors: Petkov, Gonzalez) S. Gazagnes RUG, DSSC (supervisors: Wilkinson, Koopmans, Kalantar-Nayestanagi) S. Ghosh RUG 1.0 (supervisors: Bunte, Petkov, Biehl) C. Haigh SUNDIAL (H2020 ITN) 1.0 (supervisors: Wilkinson, Trager, Petkov)

117 Bernoulli Institute Annual Report

M. Lange HS Mittweida external (supervisors: Biehl, Villmann) M. Lopez Antequera Univ. Malaga, RUG UE 0.0 (supervisors: Petkov, Gonzalez) F. Melchert IFF Magdeburg, RUG UE (supervisors: Biehl, Seiffert) M. Mohammadi SUNDIAL (H2020 ITN) (supervisors: Bunte, Petkov, Biehl) M. Straat RoSF 1.0 (supervisors: Bunte, Biehl) D. Nebel HS Mittweida external (supervisors: Biehl, Villmann) A. Nolte SUNDIAL (H2020 ITN) 1.0 (supervisors: Biehl, Petkov) G. Owomugisha Makerere University external (supervisors: Biehl, Mwebaze, Quinn) E. Schiza Univ. Cyprus external (supervisors: Petkov, Pattichis) C. Shi priv. funds 1.0 (supervisor: Petkov) E. Talavera Martinez Univ. Barcelona, RUG UE 0.0 (supervisors: Petkov, Radeva) X. Zhang priv. funds 1.0 (supervisor: Wilkinson) M. Leyva Vallina EU H2020 TrimBot 1.0 (supervisor: Petkov, Wilkinson, Strisciuglio)

118 Bernoulli Institute Annual Report

Guests in 2018: L. Alves, Univ. Sao Paulo, Brazil G. Bhanot, Rutgers University, New Jersey, USA M. Cannataro, Univ. Magna Grecia, Catanzaro, Italy M. Castejon Limas, Univ. of Leon, Spain N. Chamba, Spain A. Costa, Univ. Sao Paulo, Brazil L. Fernandez Robles, Univ. of Leon, Spain C. Gopfert,¨ Bielefeld University, Germany J.-P. Gopfert,¨ Bielefeld University, Germany Pal˚ Grandal, Univ. of Agder, Norway Ole-Christoffer Granmo, Univ. of Agder, Norway M. Kaden, Univ. of Applied Sciences Mittweida, Germany S. Kalitzin, Netherlands Institute on Epilepsy M. Lange, Univ. of Applied Sciences Mittweida, Germany E. Mwebaza, Makerere Univ. Kampala, Uganda D. Nebel, Univ. of Applied Sciences Mittweida, Germany C. Neocleous, Tech. Univ. of Cyprus C.S. Pattichis, Univ. of Cyprus S. Saralajew, Porsche AG, Weissach, Germany C.N. Schizas, Univ. of Cyprus F.-M. Schleif, Univ. of Applied Sciences Wurzburg,¨ Germany T. Villmann, Univ. of Applied Sciences Mittweida, Germany P. Tino, University of Birmingham, UK B. Bideran, University of Heidelberg, Germany A. Hung-Shuo Su, University of Oulu, Finland

119 Bernoulli Institute Annual Report

120 Bernoulli Institute Annual Report

9.1 Research Program

Our mission is to pursue high quality research in the theory, development, and application of Intelligent Systems. We communicate our results through highly renowned journals and at leading international conferences. Emphasis is put on training graduate students and guiding post-doctoral researchers in order to achieve outstanding results. In close collaborations with domain experts we address highly relevant, interdisciplinary applications of computer science in life sciences, health care, finance, robotics, animal breeding, astronomy and other areas. Our research program addresses the highly inter-connected areas of image processing and analysis, computer vision, pattern recognition, machine learning, brain-inspired computing, data science and related topics. We intend to continue developing effective and efficient methods and algorithms in the general area of Intelligent Systems. By doing so, we will participate in the grand challenge of giving computers the abilities to perceive, analyse, learn, take decisions and to enhance human creativity. Various trans-disciplinary applications from areas such as biology, medicine, life sciences in general, astronomy, robotics or the financial industry continue to provide inspiration for our work. Notably, bio-medical and health- care applications constitute an important focus for our research and a number of long-term collaborations have been established. In addition, astroinformatics has become one of our most active application areas, recently. We participate in the school of Behavioural and Cognitive Neurosciences (BCN) with the brain- inspired computing aspects of our research. We are also members of the Advanced School of Computing and Imaging (ASCI) and we contribute to its course programme. We participate in the EU H2020 robotics project TrimBot2020 and the NWO-STW project SmartBreed. We are actively involved in the EU H2020 Innovative Training Network (ITN) SUNDIAL. Furthermore, we collaborate with the Groningen Engineering Center in the Regions of Smart Factories project, a conglomerate of European industry and research facilities. We also contribute significantly to the activities of the Data Science and Systems Complexity Centre (DSSC) of the Faculty of Science and Engineering. Our group plays an instrumental role in the Intelligent Systems specialization of the MSc programme in Computer Science, by offering courses in Data Science, Computer Vision, Pattern Recognition, Neural Networks and Computational Intelligence and by providing research projects for graduate students. These graduate courses are followed also by many students of Artificial Intelligence and other programs.

Brain-inspired computing in pattern recognition and computer vision We develop models of information processing in visual cortex and use them in computer algorithms. This research is relevant for the areas of image processing, computer vision, pattern

121 Bernoulli Institute Annual Report recognition, visual perception, and computational neuroscience. Our goal is to understand how people see and to deploy principles of natural vision in computer algorithms for artificial vision. Using facts from neuroscience and visual perception, we build models of visual information processing in the brain and use them in computer simulations to obtain insights and derive practical computer vision algorithms. One example is a model of a grating cell that we developed [Petkov, Kruizinga: 1997 Biol. Cyb. 76: 83-96] and used in a texture operator [Kruizinga, Petkov: 1999 IEEE Trans. Im. Proc. 8: 1395-1407], [Grigorescu, Petkov, Kruizinga: 2002 IEEE Trans. Im. Proc. 11: 1160-1167]. By means of computer simulations we demonstrated that grating cells may play an important role in the disambiguation of edge information in early vision (texture vs. contours). Another example is our model of non-classical receptive field inhibition, also called surround suppression, in orientation selective neurons [Petkov, Westenberg: 2003 Biol. Cyb. 88: 236-246]. We demonstrated that the biological role of this inhibitory mechanism is quick pre-attentive detection of object contours and region boundaries in natural images that are rich in texture. Later, we studied general contextual modulation in vision and its de-texturizing effect [Gheorghiu et al., Vis. Res. 104, 12-23, 2014]. We proposed various contour detection algorithms that deploy this mechanism and showed that they are more effective in detecting object contours and region boundaries than traditional computer vision algorithms for edge detection [Grigorescu, Petkov, Westenberg: 2003 IEEE Trans. Im. Proc. 12: 729-739], [Grigorescu, Petkov, Westenberg: 2004 Im. Vis. Comp. 22: 609–622], [Papari, Campisi, Petkov, Neri: 2007 EURASIP J. Adv. Sig. Proc., Article ID 71828]. This work has been extended by applying gestalt principles to edge grouping [Papari, Petkov: 2008 IEEE Trans. Im. Proc. 17: 1950-1962], [Papari, Petkov: Proc. SPIE 2008, vol. 6812, art. no. 68121B]. We also studied the orientation and speed tuning properties of spatio-temporal 3D Gabor and motion energy filters with surround suppression as models of time-dependent receptive fields of simple and complex cells in primary visual cortex (V1) [Petkov, Subramanian: 2007 Biol. Cyb., 97: 423-439]. We demonstrated how these filters are related to motion detection, noise reduction, texture suppression and contour enhancement. We developed a new computational model of a simple cell that outperforms the popular Gabor function model [Azzopardi, Petkov: 2012 Biol. Cyb. 106 (3): 177-189], which was later extended by introducing inhibition [Azzopardi, Rodriguez-Sanchez, Piater, Petkov: 2014 PLOS ONE, vol. 9 (7): e98424. doi:10.1371/journal.pone.0098424, 2014 In the same line of research we modeled shape representation in areas V4 and TEO of visual cortex and applied our models to various practical problems [Azzopardi, Petkov: 2011 LNCS 6854 : 451-459], [Azzopardi, Petkov: 2012 Patt. Rec. Lett. http://dx.doi.org/10. 1016/j.patrec.2012.11.002], [Azzopardi, Petkov: 2013 IEEE Trans. PAMI 35 (2): 490-503], [Azzopardi, Petkov: 2013 LNCS 8048 : 9-16], [Azzopardi, Petkov: 2014 Frontiers

122 Bernoulli Institute Annual Report in Computational Neuroscience, vol. 8(80)], [Azzopardi, Petkov: 2014 BrainComp, LNCS, vol. 8603] We call our approach to model design combination of receptive fields (CORF). We developed an efficient implementation of the mentioned models that we call ’combination of shifted filter responses’ (COSFIRE). CORF models and COSFIRE filters are trainable as they can be configured by the automatic analysis of a user-specified prototype pattern. Recently we combined our COSFIRE trainable filters approach with CNN representations and the results will be published in 2019. Another example of our research that is inspired by psychophysical research on the human visual system is a method for the evaluation of the robustness of shape recognition algorithms to incompleteness of contours [Ghosh, Petkov: 2005 IEEE Trans. PAMI 27: 1793-1804]. Recently, we extended this line of research to the design of algorithms that deploy models of functions of the auditory system of the brain and their application to sound signal analysis. We developed an algorithm for audio feature extraction and recognition of sound events called COPE (Combination of Peaks of Energy), the design of which is inspired by the mechanism that converts sound pressure waves that reach the outer ear into neural stimuli on the auditory nerve [Strisciuglio, Vento, Petkov: 2016 BrainComp, LNCS, vol. 10087]

Image processing and computer vision We study processing operations that are fundamental for computer vision, such as contour detection. This resulted in a large survey: G. Papari and N. Petkov: Edge and line oriented contour detection: State of the art, J. Im. Vis. Comp., 29 (2-3), 2011, 79-103. Moreover, we have proposed a novel contour detector that relies on simple operations: convolutions with difference-of-Gaussian filters, blurring, shifting and multiplication [Azzopardi, Petkov (2012): LNCS 7552: 395-402]. In shape analysis we study geometrical approaches in which a feature point is characterized by the spatial arrangement of other feature points around it. The collection of local geometrical descriptors for the different feature points of an object is used as a shape characteristics of that object [Grigorescu, Petkov: 2003 IEEE Trans. Im. Proc. 12: 1274-1286]. In the same context we developed a method for the automatic construction of trainable filters that we called COSFIRE (Combination of Shifted Filter Responses) [Azzopardi, Petkov: 2013 IEEE Trans. PAMI 35 (2): 490-503], [Gecer et al., Image and Vision Computing 57: 165-174] and we demonstrated the effectiveness of this approach in various practical applications. Another direction in our work is the development of image processing operators that add artistic effects to photographic images [Papari, Petkov, Campisi: 2007 IEEE Trans. Im. Proc., 16: 2449-2662], [Papari, Petkov: 2009 IEEE Trans. Im. Proc. 18: 652-664].

123 Bernoulli Institute Annual Report

On the applications side, we collaborate with researchers from the University of Leon, Spain, in the area of automatic classification of boar spermatozoa [Sanchez, Petkov, Alegre: 2006 Cell. Mol. Biol., 52: 38-43], [Petkov, Alegre, Biehl, Sanchez: 2008 Comp. Biol. Med. 38: 461-468]. We also collaborate with the Department of Dermatology of the University Medical Center Groningen on the application of content based image retrieval and expert systems to dermatological problems [Bosman, Petkov, Jonkman: 2010 Skin Res. and Tech. 16: 109-113], [Bunte et al: 2011 Patt. Rec. 44 (9): 1892-1902], [Giotis et al: 2012 Skin Res. and Tech., 19 (1), E123-E131]. We have also proposed a method to automatically localize the vascular bifurcations in retinal images, a process that is important for the diagnosis of several cardiovascular diseases. [Azzopardi, Petkov: 2013 Pattern Recognition Letters, 34 (8): 922-933]. We have proposed a novel delineation algorithm and demonstrated that it is highly effective and efficient for the segmentation of the vessel trees in retinal fundus images [Azzopardi, Strisciuglio, Vento, Petkov: 2015 Medical Image Analysis vol.19(1), 46-57]. Recently, we have extended our delineation algorithm by inclusion of a push-pull inhibition model of some cells in the visual cortex of the brain, which improves the response in noisy and textured regions of the images [Strisciuglio, Azzopardi, Petkov: 2018, ECCVW 2018 (LNCS, volume 11131:555-565)]. Another application to health care is the system for personalized advice for schizophrenia patients that we develop with the group Distributed systems (M. Aiello) and researches from the psychiatry department of UMCG [Emerencia et al 2013 AI in Medicine 58 (1): pp. 23-36]. We participate in the EU H2020 robotics project TrimBot for the design of a gardening robot for trimming of bushes and flowers. Specifically, we are involved in the development of the computer vision system of that robot. Within the SmartBreed project, in collaboration with Wageningen University and Research and the company Topigs-Norsvin we developed a new method to automatically assign mascularity scores to pigs using imaging data.

Big data and applied machine learning In the area of big data, we apply machine learning to the analysis of data of pigs and develop methods for the optimal assignment of pigs to finishing pens. We do this in collaboration with Wageningen University and Research and the pig breeding company Topigs-Norswin. In collaboration with several UK universities we use big data from health records provided by the NHS and analyse it with machine learning methods to estimate suicide risk in patients. In collaboration with the Univ of Cyprus we work in the area of ehealth.

124 Bernoulli Institute Annual Report

Connected filters, Connectivity Theory and Segmentation Connected filters are a field of research within mathematical morphology which has seen huge developments in the last two decades. They are edge preserving operators which have found use in noise removal, texture analysis, image compression and description, and feature extraction. Research on connected operators in our group entails algorithm development (including parallelization), development of new classes of filters, applications to 2-D and 3-D medical images, and the development of new connectivity measures for these filters for increased robustness. Recently, processing of giga-pixel and tera-pixel scale images for remote sensing and astronomy has been added to this list of applications. One line of this research links to visual cortex modelling: developing morphological ana- logues of texture operators based on models of certain visual cortical cells. It is hoped these morphological counterparts will be an order of magnitude faster, whilst retaining the useful properties of the cortical cell models. Finally, fast visualization based on connected attribute filters is being explored. We have expanded this line into hyperconnected filters and attribute- space-connected filters, which increase the flexibility of perceptual groupings available, and allow dealing with overlap explicitly. Extensions to colour, or other vector images, and even tensor images are being developed, both for regular colour and multi-band processing and for diffusion tensor imaging (DTI). Applica- tions include brain imaging, astronomical imaging, remote sensing, and other applications. This research also has close links to machine learning, because machine learning techniques are becoming incorporated into the morphological filtering methods themselves. Learning Vector Quantization has been used to teach filters to enhance certain objects in images. Segmentation is a core problem in image analysis, and methods based on both simple thresh- olding methods and more advanced methods such as watersheds and deformable models are being explored. Application areas are many, but the focus lies on biomedical imaging, both macroscopic (MRI, CT) and microscopic. New application domains in astronomy are also being explored, and in particular automatic detection of astronomical sources in massive image databases.

Machine Learning and Computational Intelligence Our research program in Computational Intelligence and Machine Learning addresses three main aspects: Theoretical studies and modelling of learning processes, algorithm development, and inter-disciplinary applications of machine learning. The theoretical investigation of model situations provides important insight into machine learning processes and helps in the optimization of known training algorithms and in the

125 Bernoulli Institute Annual Report design of entirely novel, efficient concepts and training schemes. We are currently revisiting the successful statistical physics approach to the theory of machine learning. In this context we are studying typical properties of learning systems in non- stationary environments, e.g. in the present of concept drift. Moreover, we have begun to investigate thoroughly the potential benefits and limitations of rectified linear units (ReLU) in layered neural networks. This specific activation function has re-gained popularity in the Deep Learning community. However, a thourough theoretical understanding of the effect it has on the models and on the learning processes is lacking. Another major target of our work is the use of adaptive distance measures and relevance learn- ing in the context of feature-based classification. Recent examples of algorithm development include specific approaches for functional data and the analysis of complex-valued data. Practical applications of modified or newly developed methods constitute an important part of our research activities. This concerns several scientific areas, including bio-medical data analysis, bioinformatics and astronomy. Obviously, the success of these lines of research hinges on intense collaboration with experts from the respective application domain. In the following, a few examples of application oriented projects are highlighted: Within the SUNDIAL H2020 Innovative Training Network, we are applying relevance learning and related methods in the areas of feature-based galaxy classification and related problems. A very successful on-going collaboration concerns the establishment and putting forward of steroid metabolomics as a diagnosis tool in endocrinology and related areas of internal medicine. Currently, a diagnostic tool for the detection of recurrent malignancy in adrenal tumors is being developed. The approach is also being extended with respect to the dis- criminative diagnosis of different types of hypertension related disorders as well as liver diseases. The classification of FDG-PET scan brain data for the diagnosis of neurodegnerative disorders continues to be investigated together with the group of Prof. N. Leenders (Neurology, UMCG) and with the the Neuroimaging Center Groningen. This collaboration is currently being extended to researchers from Tel Aviv University in Israel. Another on-going activity in the analysis of bio-medical data concerns the detection and classification of crop plant diseases in a joint project with Makerere University in Kampala and the National Crop Research Center in Namulonge, Uganda. Our collaboration with the Fraunhofer Institute IFF in Magdeburg/Germany and the acquired expertise in the analysis of hyper-spectral and other functional data is instrumental in this context. Methodological challenges like the ones mentioned above will continue to be in the focus

126 Bernoulli Institute Annual Report of our research interests. The close contact with real world applications constantly reveals interesting problems and generates new research questions.

Learning interpretable models for interdisciplinary data analysis To date most successful machine learning techniques for the analysis of complex interdisci- plinary data predominantly use significant amounts of vectorial measurements as input to a statistical system. The domain expert knowledge is often only used in data preprocessing and the subsequently trained technique appears as a black-box, which is difficult to interpret or judge and rarely allows insight into the underlying natural process. Therefore we are interested in the principled integration of expert knowledge in our learning algorithms to tackle especially computationally challenging problems and interpretable models and visualization to involve the domain experts. In the following, a few examples of projects are highlighted: In many bio-medical applications the underlying biological process is complex and the amount of measurements is limited due to the costs and inconvenience for the patient. The classification of inborn steroidogenic disorders exhibits several computational challenges ranging from very heterogeneous measurements, systematically missing values and imbalanced classes due to the rarity of some of the diseases. These disorders primary present in the paediatric population and need to be diagnosed as early as possible, to avoid delays of lifesaving therapy and to facilitate gender allocation and surgical planning in patients with disordered sex development. Since the subject population consist mostly of neonates, and babies, we have to depend on the most non- intrisive methods of diagnosis, namely urinal measurements. Several computational challenges arise from this data ranging from very heterogeneous measurements, systematically missing values and imbalanced classes due to the rarity of some of the diseases. In close collaboration with the University of Birmingham we achieved first promising results distinguishing different conditions and healthy controls very accurately (see Figure 3). Furthermore, we investigate dimensionality reduction and visualization aiming at the preserva- tion of as much information as possible important for the current task. A wealth of dimension- ality reduction techniques for data visualization and processing has been established. However, non-parametric methods require additional effort for out-of-sample extensions, because they provide only a mapping of a given finite set of points. Furthermore, the investigation of the generalization ability with respect to new data points for such methods is unclear. We proposed a general framework for dimensionality reducing data visualization to tackle these issues by extending non-parametric dimension reduction to explicit mappings to enable direct out-of-sample extension and the investigation of the generalization error. Another example is the formulation of an information retrieval based neighborhood preservation method for visual- ization on a discretized output display. Unlike previous low-dimensional neighbor embedding methods, our formulation based on Maximum satisfiability is guaranteed to yield globally optimal visualizations, and is designed for low-dimensional visualization where evaluation and

127 Bernoulli Institute Annual Report

Healthy Prot Healthy HSD3B2 Prot HSD3B2 CYP17A2 Prot CYP17A2 CYP21A2 Prot CYP21A2 PORD Prot PORD SRD5A2 Prot SRD5A2

Figure 3: Projection of the steroidogenesis data from the 6 different classes. The domain boundaries are drawn denote our classifier decision boundaries. minimization of visualization errors are crucial. In a real-world case study for semi-supervised WLAN signal mapping we showed the utility of the proposed technique for the construction of radio maps for indoor navigation based on wireless signals. The approach is able to integrate prior/user knowledge (e.g. the position of walls and key points) in a principled way into the cost function in form of soft and hard constraints. Moreover, we are part of the Innovative Training Network SUNDIAL (SUrvey Network for Deep Imaging Analysis & Learning). The network started in 2017 and combines the expertise of Astronomy and Computer Science from nine European research groups and 5 private companies bridging the gap between the academic world and society. The aim is to develop novel algorithms to study the very large databases coming from current-day telescopes to better understand galaxy formation and evolution, convert results into commercial products and, last but not least, train 14 early stage researchers to face the interdisciplinary challenges of the next decade. Last but not least, within the Regions of Smart Factories (RoSF) project we are developing algorithms for mechanical and control engineering, concerning large-scale interconnected and production line models for the prediction of faults. Sensors measure the output performance and provide corrective feedback aiming to achieve the desired behaviour of the controlled process. We aim for task-driven analysis of dynamic systems it has the potential to contribute to the fourth industrial revolution (Industry 4.0), known in the Netherlands as Smart Industry. Digitalization should render factories more efficient and flexible using interpretable machine

128 Bernoulli Institute Annual Report learning techniques to detect individual deviations of constructed systems, predict faults in production lines and decay in infrastructure, which fits very well in the Northern Netherlands Innovation Agenda.

9.2 Overview of scientific results

Brain-inspired computing in pattern recognition We developed a new brain-inspired robust delineation operator, see publications [Strisciuglio et al. 2018].

Computer vision, audio and image processing, pattern analysis We developed a new method for camera localization in outdoor garden environments using artificial landmarks and contributed with this and other methods to the EU H2020 project Trimbot2020, see publications [Strisciuglio et al. 2018]. In collaboration with the University of Cyprus we developed a new modular domain-to-domain translation network for images, see publications [Karatsiolis et al. 2018]. Big data and applied machine learning In collaboration with Wageningen University and Research and the company Topigs-Norsvin we developed new methods for the prediction of slaughter age in pigs and the estimation of live muscle scores of pigs with RGB-D images and machine learning, see publications [Alsahaf et al. 2018]. In collaboration with several UK universities we used big data from health records provided by the NHS and analysed it with machine learning methods to estimate suicide risk in patients, see publications [del Pozo Banos et al. 2018]. In collaboration with the Univ of Cyprus we made a proposal for an ehealth based ecosystem serving national healthcare, see publications [Schiza et al. 2018].

Connected filters, Connectivity Theory and Segmentation In terms of connected filters, several improvements of attribute filtering were made, in par- ticular in terms of algorithms. An important development is the study of remote sensing data, in collaboration with the European Commission Joint Research Centre, in Ispra, Italy, and DigitalGlobe Inc., Westminster, Colorado, USA. This entails multi-scale morphological studies of huge data sets, ranging from gigapixel to terapixel scale, with the aid of Differ- ential Morphological Profiles (DMP). This type of analysis is used to detect, e.g., rubble in

129 Bernoulli Institute Annual Report images of areas affected by earthquakes, tsunamis or other disasters. We have improved the previously published versions of the DMP algorithm to improve cache coherence and reduce computational overhead, and tested scalability on shared memory machines with up to 64 cores. Further extensions include the Differential Area Profile, and the derived multi-scale levelling segmentation. Building on previous results, and using a new 64-core parallel machine obtained through funding from NWO, we are abble to perform multi-scale analysis at a rate of about 4 Gpixels per minute, on this one compute server, attaining speed ups of up to 50 on 64 cores. × This work has been published in the ISPRS International Journal of Geo-Information. An important new application is that of the Global Human Settlement Layer http:// ghslsys.jrc.ec.europa.eu/, which aims to give a complete, up-to-date overview of all human habitation on earth, and an unprecedented resolution. We are now moving to distributed memory algorithms for tera-pixel images. For this purpose a new algorithm has been devised, which changes the maximum communication and memory loads per node back from O(N) to O(√N). Apart from work on remote sensing applications, very large stitched, 360 deg images can be processed using these methods. To achieve this, the image is first split into disjoint tiles. Instead of attempting to build a single component tree for the entire image, we build a forest of component trees, one for each tile. Data concerning structures touching the tile boundaries are then exchanged, to modify each tree in such a way that filtering them yields exactly the same result as filtering the entire tree. This approach works well for low dynamic range images (8-12 bits per pixel). The first MPI implementation of area filters using this approach has been made, and we have presented the results at the International Symposium on Mathematical Morphology (ISMM) 2017 in Fontainebleau, France, and at ICT-Open 2017. Further extensions have been made in to this work to include area profiles and multi-scale levelling segmentations on images up to 163 Gpixel. Speed-ups of up to 103 on 128 MPI tasks on 128 nodes were achieved. We are working on increasing the flexibility of this algorithm to allow irregular grids of tiles, and processing of streaming data from satellites or aircraft, as well as extensions to images too big to fit into the memory of the current clusters. Extensions to 3D are also being developed. Our research in the field of astronomical imaging and object detection has also lead to insights into shortcomings of an important astronomical image processing tool: Source Extractor. This tool is essentially based on a heavily discretized “poor-man’s” Max-Tree. We improved this by incorporating a real Max-Tree implementation, and implementing a new strategy for Max-Tree filtering called bi-variate statistical attribute thresholding. In methodology, we determine a threshold based on one attribute based on a statistical test on the likelihood that the feature detected is due to noise, as a function of the area of the feature. In this case the user needs to supply a p-value which determines the probability that a noise feature is rated as a true

130 Bernoulli Institute Annual Report object. For large objects, a small deviation of, e.g., the image power within it over the local background suffices to rate as a detection, whereas a a small object needs to have a higher image power. The results show great promise, as the sensitivity of the method to extended sources seems to be an order of magnitude better than SExtractor, whilst retaining excellent sensitivity for compact sources, and negligible false positive count, even at a conservative 6 p-value of 10− . The initial results were presented at the PRIA-2013 conference and at ICT-Open 2013. Further improvements were presented at the ISMM-2015 in Reykjavik, and a journal paper was published this your in Mathematical Morphology – Theory and Application. Within the SUNDIAL project, we are extending this research to a full comparison of faint object detection methods in astronomical images, including NoiseChisel, SExtractor, and the most recently published ProFound, on a larger set of deeper images which provide a sterner test of these methods than the Sloan Digital Sky Survey used before. The aim is to develop statistical tests for attribute filters which are less sensitive to the presence of outliers (stars). In astronomy, most data is either floating point or integer with very high dynamic range. The existing parallel algorithm does not handle these well at all. This means that we must develop a parallel Max-Tree algorithm for floatingpoint and other high-dynamic-range data. The new algorithm has been developed and is a combination of a bottom-up flooding approach to compute a coarse approximation Max-Tree, followed by top-down refinement phase based on the union-find methods, thus combining the two leading schools of thought on these algorithms into a single “diplomatic” algorithm. Speed-up of up 40 have been achieved on 64 cores, × which translates to a 30-fold speed increase with respect to the fastest sequential algorithm. This work has been published in IEEE trans. Pattern Anal. Mach. Intell. in 2018. As of this writing, there is no published algorithm for distributed computation of these attribute filters on extreme dynamic range images (> 16 bits per pixel). A collaboration from Paris and Juelich do have a working prototype, but its performance should be improved. We are starting a collaboration to develop such a tool. In collaboration with the KVI-CART institute a new method based on attribute-space connected filters has been developed for tracking subatomic particles through a complex series of detectors in the PANDA experiment. Rather than using (very costly) circular Hough transforms we perform connected openings in an orientation-based attribute space. The results of the first proof-of-concept algorithm show that the new algorithm can detect over 80% of tracks correctly. These results were published in Mathematical Morphology – Theory and Application. An improved method, with lower computational complexity (O(N log N) rather than O(N 2)), and increased detection performance has been developed (roughly 90% correct segmentation). These results have been submitted to IEEE Transaction on Image Processing. In the process of this work, a new graph formalism for attribute-space connected filters was developed, which leads to a significant reduction in memory load. A further extension of this has been

131 Bernoulli Institute Annual Report developed in collaboration with prof A.S. Chowdhury of Jadavpur University, Kolkata, India. and submitted to ICIP 2018. Finally, A new segmentation method based on alpha-trees was developed for semi-automatic segmentation of images of anatomical specimens, by visiting MSc student Wilbert Tabone. These results have also been submitted to

Machine learning and Computational Intelligence Here we highlight a few key results by referring to example journal articles which appeared in 2018. A number of recent publications addressed the theoretical investigation of machine learning paradigms in model situations. As a first example, the article Statistical Mechanics of On-line Learning Under Concept Drift presents theoretical results obtained in collaboration with researchers from Bielefeld University [Straat et al., Entropy 20 (10, article number 775, 2018]. The possibility to directly optimize a classifier’s Receiver Operating Characteristics (ROC) in the learning process is studied in [Villmann et al., Computational Statistics, 33(3): 1173-1194, 2018]. Several publications address the development of suitable training algorithms and classifiers for functional data, such as spectral data or time series. The use of complex representations is discussed in the paper Learning Vector Quantization and Relevances in Complex Coefficient Space by Straat et al. [Neural Computing and Appplications, 15 pages, online, 2018]. Our recent machine learning based analysis of astronomical data in the context of galaxy classification was published by Nolte et al. [Neurocomputing, 342: 172-190, 2019, on-line 2018]. Several journal publications emerged from successful local and international collaborations. For instance, the paper Effect estimate comparison between the prescription sequence symmetry analysis (PSSA) and parallel group study designs by Idema et al. appeared in [Plos ONE, 13(12): e0208389, 2018]. In addition, a number of refereed conference contributions and book chapters were published in 2018 in the context of the above mentioned areas of activity.

Learning interpretable models for interdisciplinary data analysis Here we highlight some results and methods published in 2018 with respect to interpretable models and interdisciplinary data analysis.

132 Bernoulli Institute Annual Report

Figure 4: Distribution of stars in the Gaia DR1 survey (left panel) and known GCs and detected candidates from window A (right panel).

Many interesting structures are buried inside large astronomical databases. One such structure is a Globular Clusters (GC), a spherically shaped group of stars bound by gravity with high stellar densities towards its center. It is assumed that its stars are formed from a single cloud of interstellar gas and dust. Therefore we suspect that stars of a GC are formed at the same time and have the same chemical composition, which makes them good candidates to study stellar evolutions. They are for example used to test stellar evolution codes. We investigated different strategies to find such spherical dense group of stars automatically in the Gaia DR1 survey. Figure 4 shows the Halo of the Milky Way investigated, the position of known GCs and candidates resembling spherically shaped stellar densities. Results were presentedIn Frontiers in Artificial Intelligence and Applications (FAIA), the European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN 2018) and published in the journal Neurocomputing 2019. Further publications approach interdisciplinary bio-medical data analysis including classi- fication of inborn steroidogenic disorders including the poster “Steroid metabolomics: A powerful technique for differentiating inborn disorders of steroidogenesis” presented at the 17thInternational Congress on Hormonal Steroids and Hormones & Cancer (ICHSCH) 2018, as well as publications in Endocrine Abstracts. Another important contribution for the princi- pled inclusion of expert knowledge into machine learning has been published in the Journal of Theoretical Biology titled “Learning pharmacokinetic models for in vivo glucocorticoid activation”. For the analysis of time series data one can take a purely data-driven machine learning approach or alternatively fit mechanistic models. To take advantage of the predictive power of machine learning and the explanatory power of mechanistic models, we propose a latent variable mixture model for learning clusters of mechanistic models. The proposed

133 Bernoulli Institute Annual Report strategy automatically constructs different population models that are not based on prior knowledge or experimental design, but result naturally as mixture component models of the global latent variable mixture model. In the context of Smart Industry we presented a project in cooperation with a local e-commerce company: “Efficient learning of email similarities for customer support” at the European Symposium on Artificial Neural Networks (ESANN) 2019. Several journal and conference publications are currently under review or in press which indicate that our research activities continue to yield significant output.

134 Bernoulli Institute Annual Report

9.3 Research subjects

A. Alsahaf: Machine learning for animal breeding. L. Alves: Applications of Learning Vector Quantization. M. Babai: (hyper)Connected morphology for particle track extraction. M. Biehl: Machine learning, statistical physics of learning, distance-based methods, life science and other applications. K. Bunte: Machine learning, interpretable models, mechanistic machine learning using princi- pled integration of expert knowledge, interdisciplinary data analysis, dimensionality reduction and visualization. D. Fernandez´ Chaves: Computer vision for robotics. S. Gazagnes: Radio-astronomical data analysis using (hyper)connected filters. S. Ghosh: Bio-medical data analysis and computer aided diagnosis C. Haigh: Optical astronomical data analysis using (hyper)connected filters. M. Lange: Information theoretical aspects of prototype based learning. M. Leyva Vallina: Computer vision for robotics. Guo Li: Signal processing. M. Lopez Antequera: Pattern recognition for robotics. F. Melchert: Classification of hyperspectral images and other functional data. M. Mohammadi: Machine learning for astronomical data analysis and astroinformatics D. Nebel: Learning from relational data, rejection mechanism for outliers. A. Nolte: Machine learning techniques in astroinformatics G. Owomugisha: Crop plant disease detection in images and spectral data. N. Petkov: Pattern recognition, machine learning, data analytics and applications. Samira Rezaie: Analysis of radio-astronomy data E. Schiza: Electronic patient file. C. Shi: Inhibition in trainable feature detectors. M. Straat: Smart Factory applications, theory of machine learning N. Strisciuglio: Computer vision for robotics, machine learning, pattern recognition. E. Talavera Martinez: Life-logging with a camera. M.H.F. Wilkinson: Morphology, connectivity, biomedical imaging. X. Zhang: Hyperspectral connected filtering for remote sensing.

135 Bernoulli Institute Annual Report

9.4 Publications

Dissertations

Eirini C. Schiza, An eHealth Driven National Healthcare Ecosystem, Supervisors: • Prof. N. Petkov (1st promotor), Prof. C.S. Pattichis (Univ. of Cyprus, 2nd promotor), Faculty of Science and Engineering, University of Groningen, , 22 June 2018, ISBN: 978-94-034-0605-3, ISBN ebook: 978-94-034-0604-6, 185 pages.

Edited volumes

R. Smedinga and M. Biehl (eds.), 15th SC@RUG 2017-2018, Proc. of the Student • Colloquium Computer Science, University of Groningen (ISBN 978-94-034-0737-1), 2018, 128 pages.

N. Petkov, N. Strisciuglio, C. M Travieso-Gonzalez´ (eds.), Applications of Intelligent • Systems: Proceedings of the 1st International APPIS Conference 2018, In the series Frontiers of Artificial Intelligence and Applications (IOS Press), Vol 310, ISBN 978-1- 61499-928-7 (print), ISBN 978-1-61499-929-4 (online).

Leszek J Chmielewski, Ryszard Kozera, Arkadiusz Orłowski, Konrad Wojciechowski, • Alfred M Bruckstein, Nicolai Petkov: Computer Vision and Graphics: International Conference, ICCVG 2018, Warsaw, Poland, September 17-19, 2018, Proceedings. LNCS, Vol. 11114, 2018

L. Onetto, K. Bunte, and F.-M. Schleif. “Advances in artificial neural networks, machine • learning and computational intelligence Selected papers from the 26th European Sym- posium on Artificial Neural Networks (ESANN 2018)”. In: Neurocomputing (2018), pp. 1–2.

Articles in scientific journals

K. Bunte, D. J. Smith, M. J. Chappell, Z. K. Hassan-Smith, J. W. Tomlinson, W. Arlt, • and P. Tino. Learning Pharmacokinetic Models for in vivo Glucocorticoid Activation. In: Journal of Theoretical Biology, 2018, Volume 455 (Oct. 2018), pp. 222–231.

D.L. Idema, Y. Wang, M. Biehl, P.L. Horvatovich, E. Hak. Effect estimate comparison • between the prescription sequence symmetry analysis (PSSA) and parallel group study designs: A systematic review. PLoS ONE, 13(12), 2018, e0208389 (open access)

136 Bernoulli Institute Annual Report

M. Straat, F. Abadi, C. Gopfert,¨ B. Hammer, M. Biehl. Statistical Mechanics of On-Line • Learning Under Concept Drift. Entropy, 20(10), 2018, Art. 775 (open access) T. Villmann, M. Kaden, W. Herrmann, M. Biehl. Learning Vector Quantization classi- • fiers for ROC-optimization. Computational Statistics, 33(3), 2018, 1173-1194. Ahmad Alsahaf, George Azzopardi, Bart Ducro, Egiel Hanenberg, Roel F Veerkamp, • Nicolai Petkov: Prediction of slaughter age in pigs and assessment of the predictive value of phenotypic and genetic information using random forest. Journal of Animal Science, Vol. 96, Issue 12, pages 4935-4943, 2018. Marcos del Pozo Banos, Carlos M Travieso, Kate Loxton, Nicolai Petkov, Damon • Berridge, Keith Lloyd, Caroline Jones, Sarah Spencer, Ann John: Combining Artificial Neural Networks, Routine Health Records and Suicide Risk Estimation. International Journal of Population Data Science, Vol. 3, Issue 4, 2018. Eirini C Schiza, Theodoros Kyprianou, Nicolai Petkov, Christos N Schizas: Proposal for • an eHealth Based Ecosystem Serving National Healthcare. IEEE journal of biomedical and health informatics, 2018. Marcos DelPozo-Banos, Ann John, Nicolai Petkov, Damon Mark Berridge, Kate South- • ern, Keith LLoyd, Caroline Jones, Sarah Spencer, Carlos Manuel Travieso: Using neural networks with routine health records to identify suicide risk: feasibility study. JMIR mental health, Vol. 5, Issue 2, 2018, pages: e10144. C. Shi, G. Azzopardi, D. Zillikens, E. Schmidt, G.F.H. Diercksr, J. Guo, J.M. Meijer, • M. Jonkman, N. Petkov, Detection of u-serrated patterns in direct immunofluorescence images of autoimmune bullous diseases by inhibition-augmented COSFIRE filters, International Journal of Medical Informatics, 2018. Nicola Strisciuglio, Leranring audio and image representations with bio-inspired train- • able featrute extractors, ELCVIA Electronic Letters on Computer Vision and Image Analysis 16 (2), pages 17-20

Articles in conference proceedings

M. Mohammadi, N. Petkov, R. Peletier, P. Bibiloni, and K. Bunte. Detection of Glob- • ular Clusters in the Halo of Milky Way. In: Frontiers in Artificial Intelligence and Applications (FAIA). Vol. 310. 2018, pp. 70–78. M. Mohammadi, R. F. Peletier, F. Schleif, N. Petkov, and K. Bunte. Globular cluster • detection in the Gaia survey. In: Proc. of the 26th European Symposium on Artificial

137 Bernoulli Institute Annual Report

Neural Networks (ESANN). Ed. by M. Verleysen. D-facto Publications, 2018, pp. 327–332.

M. Biehl, K. Bunte, G. Longo, P. Tino. Machine Learning and Data Analysis in • Astroinformatics. In: M. Verleysen (editor), Proc. of the 26th European Symposium on Artificial Neural Networks ESANN 2018, Bruges, Belgium, Ciaco-i6doc.com, 2018, 307-314.

A. Nolte, L. Wang, M. Biehl. Prototype-based analysis of GAMA galaxy catalogue data • In: M. Verleysen (editor), Proc. of the 26th European Symposium on Artificial Neural Networks ESANN 2018, Bruges, Belgium, Ciaco-i6doc.com, 2018, 339-344.

G. Bani, U. Seiffert, M. Biehl, F. Melchert Adaptive Basis Functions for Prototype-based • Classification of Functional Data In: Proc. of the 12th Intl. Workshop on Self-Organizing Maps, Learning Vector Quantization, Clustering and Visualization (WSOM+) Nancy/France, IEEE Xplore, 8 pages, 2017.

G. Owomugisha, F. Melchert, E. Mwebaze, J.A. Quinn, M. Biehl. Machine Learning for • diagnosis of disease in plants using spectral data. In: Proc. of the Intl. Conf. Artificial Intelligence ICAI, CSREA Press, 2018, 9-15.

R. van Veen, L. Talavera Martinez, R.V. Kogan, S.K. Meles, D. Mudali, J.B.T.M. • Roerdink, F. Massa, M. Grazzini, J.A. Obeso, M.C. Rodriguez-Oroz, K.L. Leenders, R.J. Renken, J.J.G. de Vries, M. Biehl. Machine Learning Based Analysis of FDG- PET Image Data for the Diagnosis of Neurodegenerative Diseases. In: N. Petkov, N. Strisciuglio, C. Travieso-Gonzalez (eds.), Application of Intelligent Systems (APPIS 2018), IOS Press, Frontiers in Artificial Intelligence and Applications, Vol 310, 2018, 280-289.

A Alsahaf, G Azzopardi, B Ducro, RF Veerkamp, N Petkov, Predicting Slaughter • Weight in Pigs with Regression Tree Ensembles, In: N. Petkov, N. Strisciuglio, C. Travieso-Gonzalez (eds.), Application of Intelligent Systems (APPIS 2018), IOS Press, Frontiers in Artificial Intelligence and Applications, Vol 310, 2018, 1-9.

Ahmad Alsahaf, George Azzopardi, Bart Ducro, Roel F Veerkamp, Nicolai Petkov: • Assigning pigs to uniform target weight groups using machine learning. Proceedings of the World Congress on Genetics Applied to Livestock Production, 11.112, 10 pages, 2018.

Nicola Strisciuglio, George Azzopardi, Nicolai Petkov: Brain-inspired robust delin- • eation operator. European Conference on Computer Vision (ECCV) Workshops, 1st

138 Bernoulli Institute Annual Report

Workshop on Brain-Driven Computer Vision, LNCS, volume 11131, pages 555-565, 2018.

Nicola Strisciuglio, Radim Tylecek, Michael Blaich, Nicolai Petkov, Peter Biber, Jochen • Hemming, Eldert van Henten, Torsten Sattler, Marc Pollefeys, Theo Gevers, Thomas Brox, Robert B Fisher: Trimbot2020: an outdoor robot for automatic gardening. ISR 2018; 50th International Symposium on Robotics (publisher: VDE), pages 1-6, 2018.

Nicola Strisciuglio, Mar´ıa Leyva Vallina, Nicolai Petkov, Rafael Muu˜oz´ Salinas: Camera • Localization in Outdoor Garden Environments Using Artificial Landmarks. 2018 IEEE International Work Conference on Bioinspired Intelligence (IWOBI), pages 1-6, 2018.

Savvas Karatsiolis, Christos N Schizas, Nicolai Petkov: Modular Domain-to-Domain • Translation Network. Proc. International Conference on Artificial Neural Networks ICANN, Artificial Neural Networks and Machine Learning, Lecture Notes in Computer Science, vol 11141, pages 425-435, 2018.

J. Buhagiar, N. Strisciuglio, N. Petkov and G. Azzopardi, Automatic Segmentation • of Indoor and Outdoor Scenes from Visual Lifelogging., Applications of Intelligent Systems Proceedings of the 1st International APPIS Conference 2018, Frontiers in Artificial Intelligence and Applications 310 (2018), 194-202 . IOS Press, Amsterdam.

N. Strisciuglio, G. Azzopardi, N. Petkov, Robust curvilinear detection operator, ECCVW • Proceedings, in print, 2018

Other publications

E. S. Baranowski, S. Ghosh, C. H. Shackleton, A. E. Taylor, B. A. Hughes, M. Biehl, T. • Guran, K. Bunte, et al. Steroid Metabolomics: A Powerful Technique for Differentiating Inborn Disorders of Steroidogenesis. Poster presented at the 17thInternational Congress on Hormonal Steroids and Hormones & Cancer (ICHSCH) 2018.

K. Taxis, J. de Boer, H.G. van der Meer, M. Biehl. Reducing the drug burden index - A • post hoc analysis of a randomised controlled trial using machine learning (Abstract). Pharmacoepidemiology and Drug Safety 27, 2018, 504.

A. Moolla, A. Taylor, L. Gilligan, J. De Boer, D. Pavlov, B. Hughes, Z. Hassan-Smith, • M. Armstrong, P. Newsome, T. Shah, L. van Gaal, A. Verrijken, S. Francque, J. Grove, N. Guha, G. Aithal, E. Barnes, W. Arlt, M. Biehl, J. Tomlinson. Staging of non-alcoholic fatty liver disease through LC-MS/MS analysis of the urinary steroid metabolome. Endocrine Abstracts, 59, 2018, OC3.3.

139 Bernoulli Institute Annual Report

I. Bancos, A. Taylor, V. Chortis, A. Sitch, K. Lang, A. Prete, M. Terzolo, M. Fassnacht, • M. Quinkler, D. Kastelan, D. Vassiliadi, F. Beuschlein, U. Ambroziak, M. Biehl, J. Deeks, W. Arlt. Urine steroid metabolomics as a diagnostic tool for detection of adrenocortical malignancy - a prospective test validation study. Endocrine Abstracts 56, 2018 OC72. T. Villmann, J.R.D. Ravichandran, S. Saralajew, M. Biehl. Dropout in Learning Vec- • tor Quantization Networks for Regularized Learning and Classification Confidence Estimation. Machine Learning Reports MLR-01-2018, 2018, 15-21. M. Biehl. The statistical physics of learning in a nutshell (Abstract). Machine Learning • Reports MLR-01-2018, 2018, 23. Ahmad Alsahaf, G Azzopardi, N Petkov, Estimation of live muscle scores of pigs with • RGB-D images and machine learning, Scientific Symposium FAIR Data Sciences for Green Life Sciences, 11 Dec 2018. DOI: https://doi.org/10.18174/FAIRdata2018.16281. A. Bhole, M. Biehl, G. Azzopardi, Automatic recognition of Holstein cattle using • non-invasive computer vision approach, FAIR 2018, Wageningen

9.5 External funding and collaboration

External funding

Godliver Owomugisha is an external PhD candidate from Makerere University, Uganda, supported by the Melinda and Bill Gates Foundation. F. Melchert is a PhD candidate in the framework of the ’sandwich’ program of the Graduate School Groningen, supported by the Fraunhofer Institute IFF in Magdeburg, Germany. C. Haigh, A. Nolte and M. Mohamadi are PhD students supported within the SUNDIAL Innovative Training Network (EU H2020 ITN). M. Straat is a PhD candidate in the Regions of Smart Factories (RoSF) project. N. Strisciuglio is a postdoc in the EU funded H2020 TrimBot project (2016-2019). M. Leyva Vallina is a PhD student financed from the same project. A. Alsahaf is a PhD student in the NWO-STW funded SmartBreed project (2016-2019). Eirini Schiza from Cyprus is enrolled as an external PhD student in Groningen and is jointly supervised by Nicolai Petkov (RuG) and C. Schizas and C. Pattichis (Univ. of Cyprus).

140 Bernoulli Institute Annual Report

Manuel Lopez Antequera and David Chavez are PhD students in the framework of a ’sandwich’ program of the graduate school. In a shared funding scheme, they spends half of the time at the University of Groningen and the other half at the University of Malaga and will obtain a double PhD degree from the two universities. They are supervised by Nicolai Petkov and Javier Gonzalez from the University of Malaga. Estefania Talavera Martinez is a PhD student in the framework of a ’sandwich’ program of the graduate school. In a shared funding scheme, she spends half of the time at the University of Groningen and the other half at the University of Barcelona and will obtain a double PhD degree from the two universities. She is supervised by Nicolai Petkov and Petia Radeva from the University of Barcelona. Chenyu (Astone) Shi is a PhD student at the group from the People Republic of China. Guo Li is a visitor for one year with financing from the PR of China. In 2018 we participated in the preparation of a H2020 project proposal AgriWare with coordinator Politecnico di Milano. The requested support was not granted. We are considering resubmission in a next round.

External collaboration

Biehl collaborates with Barbara Hammer’s group at CITEC, Bielefeld University, in the context of the theory and application of prototype based clustering, classification, and visualization of high-dimensional data. The development of novel algorithms and alternative distance measures is also studied in collaboration with Thomas Villmann in Mittweida, Germany. The application of LVQ in steroid metabolomics is in the center of ongoing collaborations with Wiebke Arlt from the University of Birmingham and with Jeremy Tomlinson at the University of Oxford, UK. Biehl furthermore collaborates with Gyan Bhanot (Rutgers University, New Jersey) in the context of genomics and proteomics data analysis. The application of machine learning in the context of life science applications is also the main topic of the collaboration with former PhD student Ernest Mwebaze and with John Quinn at Makerere University in Kampala, Uganda. In collaboration with Prof. Udo Seiffert, Fraunhofer Institute IFF Magdeburg, Germany, Biehl investigates the clustering and classification of hyperspectral data representing samples of organic material. Biehl has active Erasmus agreements with Profs. Hammer (Bielefeld), Villmann (Mittweida), and Schleif (Wurzburg)¨ .

141 Bernoulli Institute Annual Report

K. Bunte collaborates with Peter Tino from the University of Birmingham, UK in the context of principled integration of expert knowledge and modeling. Furthermore, the collaboration with Birmingham includes Wiebke Arlt from the Medical School for computer aided diagnosis of metabolic syndromes and David Smith from Applied Mathematics working on mathematical modeling for bio-medical applications. The development of model based learning is in the center of an ongoing project involving Michael Chappell from the School of Engineering at the University of Warwick, UK. Due to the intensive collaboration Kerstin Bunte is also Honorary Fellow in both the University of Birmingham and the University of Warwick. In the context of big data, dimensionality reduction and visualization Kerstin Bunte fosters contact to Barbara Hammer from the University of Bielefeld, Germany, as well as Michel Verleysen and John Aldo Lee from the Universite´ catholique de Louvain, Belgium. In Groningen the innovative training network SUNDIAL, a collaborative project between several astronomy and computer science European research groups, coordinated by Reynier Peletier from the Kapteyn Institute has started in 2017. At the UMCG she started a project for the analysis of Asthma data with the Pulmonologist Maarten van den Berge, which is an expert for airway diseases such as Asthma and COPD. Furthermore, K. Bunte is part of a new project for navigation, control and video analysis together with M. Wilkinson and B. Jayawardhana from the Faculty of Science and Engineering in collaboration with an industrial partner. Bunte has initiated an Erasmus+ agreement between the University of Birmingham, as well as the University of Gent and the University of Groningen.

N. Petkov collaborates with the Department of Dermatology of UMCG on the application of content based image retrieval and expert systems to dermatologic problems. He also collab- orates with N. Jansonius from the Ophthalmology department of UMCG on the processing of retinal fundus images for large scale screening of glaucoma. N. Petkov and J. Gonzalez Jimenez from the University of Malaga jointly supervise the ’sandwich’ PhD students M. Lopez Antequera and D. Fernandez´ Chavez. N. Petkov collaborates with M. Vento and A. Saggese from the University of Salerno in computer vision and audio processing. In 2018, he hosted three graduate Erasmus exchange students from Salerno. N. Petkov, Ch. Schizas and C. Pattichis from the Univ. of Cyprus collaborate in the area of ’electronic patient file’ and e-health and jointly supervise the PhD student E. Schiza. N. Petkov collaborates with E. Alegre, M. Castejon and L. Fernandez Robles from the University of Leon; they jointly prepared an EU project proposal. He also collaborates with L. Sanchez from that university. N. Petkov collaborates with P. Radeva from the Univ. of Barcelona; together they supervise the ’sandwich’ PhD student E. Talavera Martinez. N. Petkov participates in the EU project TrimBot2020 and collaborates with scientists from the Universities of Edinburgh, Freiburg, Amsterdam, Wageningen, ETH Zurich and the companies Bosch and DLO. N. Petkov participates in the NWO-STW project SmartBreed

142 Bernoulli Institute Annual Report and collaborates with scientists from the University of Wageningen (Bart Ducro and Roel Veerkamp) and various animal breeding companies. N. Petkov and N. Strisciuglio collaborate with Carlos Travieso Gonzalez from the Univ. of Las Palmas de Gran Canaria for the organisation of a series of conferences APPIS on the application of intelligent systems. In January 2018 they organised the conference APPIS 2018. N. Petkov collaborates with L. Grandinetti (Cosenza), K. Amunts (Duesseldorf) and T. Lippert (Juelich) for the organisation of the series of workshops BrainComp. In 2018, N. Petkov was visited by C.N. Schizas, C. Neocleous and C.S. Pattichis from Univ of Cyprus. In July 2018 N. Petkov participated in the PhD (evaluation and defence) committee of Maya Aghaei (Univ. Barcelona). Nicolai Petkov gave an invited keynote lecture at the 19th International Workshop on Combi- natorial Image Analysis IWCIA 2018 in Porto, Portugal. Nicolai Petkov gave an invited keynote lecture at the 13th International Joint Symposium on Artificial Intelligence and Natural Language Processing iSAI-NLP 2018 and the 13th International Conference on Knowledge, Information and Creativity Support Systems KICSS 2018 in Pattaya, Thailand. Nicolai Petkov gave an invited plenary lecture at the International Conference on Computer Vision and Graphics ICCVG 2018 in Warsaw, Poland. Ole-Christoffer Granmo and Pal˚ Grandal from the University of Agder, Norway, visited our group in October. Ole-Christoffer gave two talks in the colloquium of the Computer Science department. Aleke Nolte has received the Best Poster Award at the NWO Symposium on Applied Com- putational Sciences (ACOS), in Eindhoven, The Netherlands, for the contribution Galaxy classification: a machine learning analysis of GAMA catalogue data. Our alumna Laura Fernandez Robles received the first prize for the best PhD thesis in the area of Computer Vision in Spain. Laura defended her thesis at the RUG in December 2016 and at the University in Leon in February 2017 and obtained the PhD degree from both universities. On May 4 we organised the 13th International Workshop on Intelligent Systems in Allers- maborg. Our guests were Mario Cannataro (Univ Magna Grecia, Catanzaro) and Stiliyan Kalitzin (Netherlands Institute on Epilepsy). We organised the 1st International Conference on Applications of Intelligent Systems APPIS 2018, 10-12 January 2018, Las Palmas de Gran Canaria. Conference chairs were N. Petkov, N.

143 Bernoulli Institute Annual Report

Strisciuglio and C. Travieso (ULPGC), supported by M. Leyva and E. Talavera. M. Biehl gave an invited plenary lecture and a tutorial. We organised a meeting of the EU H2020 project Trimbot, 8-9 January 2018, Las Palmas de Gran Canaria. In 2018, Petkov was Erasmus+ lecturer at the University of Cyprus.

Wilkinson collaborates with the Kapteyn Institute, to explore the use of hyperconnected filters and especially vector-attribute filters in automatic source extraction in image data bases, which resulted in the HyperGAMMA project funded by NWO-EW. This has been extended to a European collaboration in the form of the SUNDIAL ITN. This also includes further collaboration with H. Talbot and L. Najman (ESIEE, Paris) on mathematical morphology. A collaboration of biomedical imaging has been set up with K.E. Purnama and T.A. Sardjono of the ITS in Surabaya, Indonesia. A very important collaboration is that with Martino Pesaresi, Pierre Soille of the European Commission Joint Research Centre, Ispra, Italy, and Kostas Stamatiou of DigitalGlobe In. Westminster, Colorado, USA, on automated concurrent image analysis for tera-scale images in remote sensing. The main aim is to speed up automatic image analysis for rapid assessment of changing situations. One application is in aiding disaster relief after earthquakes and other natural and man-made disasters. Evaluating land usage and its changes in time is another. Finally, a collaboration with Ananda Chowdhury from Jadavpur University Kolkata, India has been set up, on graph-based methods for mathematical morphology.

9.6 Further information

N. Petkov is member of the editorial boards of J. of Image and Vision Computing (Elsevier), J. of Neural, Parallel and Scientific Computations (Dynamic Publ.), Int. J. for Computational Vision and Biomechanics (Serials Publ.) Int. J. of Hybrid Intelligent Systems (IOS Press), and Int. J. of Fuzzy Logic and Intelligent Systems. He was member of the PCs of several conferences. Biehl is associate editor of the journals Pattern Recognition and Neural Processing Letters. He is member of the scientific committee of the European Symposium on Artificial Neural Networks (ESANN) conference series and has served as Program Committee member of numerous other international conferences. He co-edited a special issue of the journal Neuro- computing, presenting selected contributions to ESANN 2017: Advancees in artificial neural networks, machine learning and computational intelligence, which appeared in 2018. Biehl is a founding member of the CIID (Inst. for Computational Intelligence and Intelligent Data Analysis e.V., Mittweida).

144 Bernoulli Institute Annual Report

Biehl presented an invited keynote talk at the Smart Factory OWL in Lemgo/Germany, at the ”6th Yearly Colloquium on Image Processing in Automation”, November 20, 2018. Biehl also presented invited lectures on Supervised Learning at the 30th Canary Islands Winter School on Astrophyiscs, La Laguna, Tenerife/Spain, in November 2018. Bunte and Biehl co-organized a special session: Machine Learning and Data Analysis in Astroinformatics at the European Symposium on Artificial Neural Networks (ESANN) in Brugge/Belgium, April 2018. K. Bunte is member of the following scientific organizations and committees: Association for the Advancement of Artificial Intelligence (AAAI) since 2014; Program committee of International Joint Conference on Neural Networks (IJCNN) since 2017; Program committee of International Conference on Computer Analysis of Images and Patterns (CAIP) since 2015; Program committee of the New Challenges in Neural Computation (NC2) workshop in connection to German Conference on Pattern Recognition (GCPR) since 2014; She was invited to participate in several Dagstuhl seminars at the Leibniz Center for Informatics, Schloss Dagstuhl/Germany: seminar 17332: Scalable Set Visualizations, seminar 16261: Integration of Expert Knowledge for Interpretable Models in Biomedical Data Analysis, seminar 15101: Bridging Information Visualization with Machine Learning, seminar 12081: Inform. Visualization, Visual Data Mining and Machine Learning, seminar 11341: Learning in the context of very high dimensional data, seminar 09081: Similarity-based learning on structures.

Wilkinson is co-chair of the International Association for Pattern Recognition (IAPR) TC-18 on Discrete Geometry and Mathematical Morphology. He is (technical) programme committee member of several conferences, including the International Conference on Image Processing 2017, and the International Symposium on Mathematical Morphology 2017. He is on the steering committee for the International Symposium on Mathematical Morphology conferences. He has given tutorials on connected filters and connectivity at various national and international conferences, and organized the course on Advanced Morphological Filters for the national research school Advanced School for Computing and Imaging (ASCI). He was invited as IAPR Distinguished Speaker at the Discrete Geometry for Computer Imagery 2017 conference in Vienna. Finally, he is member of the cluster Computer Vision Noord Nederland, a consortium of companies and academia seeking to stimulate the field of computer vision in the Northern Netherlands.

Software

- We provide a Matlab toolbox Relevance and matrix adaptation in LVQ (GRLVQ, GM- LVQ, LiRaMLVQ) at http://matlabserver.cs.rug.nl/gmlvqweb/web/.

145 Bernoulli Institute Annual Report

- A beginner’s toolbox in Matlab, GMLVQ demo code, is provided and regularly updated at http://www.cs.rug.nl/˜biehl/gmlvq - Matlab implementations for Gabor filters, COSFIRE keypoint detector, and CORF contour detector are available on http://matlabserver.cs.rug.nl/

146 Bernoulli Institute Annual Report

10 Scientific Visualization and Computer Graphics

Group leader: Prof. dr. J.B.T.M. Roerdink

Tenured staff (BI members) source fte Prof. dr. J.B.T.M. Roerdink RUG 1.0 Prof. dr. A.C. Telea RUG 1.0

Tenure track assistant professors Dr. J. Kosinka RUG 1.0

PhD students L. Amabili NWO/eScience 1.0 (supervisor: Kosinka, Roerdink) P. Barendrecht RUG 1.0 (supervisor: Kosinka, Roerdink) X. Chen (since 10-2018) CSC 1.0 (supervisor: Telea, Kosinka) M. Espadoto FAPESP external (supervisor: Telea) G.J. Hettinga Dutch scholarship 1.0 (supervisor: Kosinka, Telea) C. Ji (until 9-2018) CSC external (supervisor: Roerdink) Y. Kim EU Cofund 1.0 (supervisor: Telea, Trager, Roerdink) J.F. Kruiger (until 16-9-2018) H2020-SESAR 1.0 (supervisor: Telea) L. Pagliosa FAPESP external (supervisor: Telea) M. C. Rodriguez FAPESP external (supervisor: Telea) E. Vernier CNPq external (supervisor: Telea, Roerdink) J. Wang (since 10-2018) CSC 1.0 (supervisor: Telea, Kosinka) X. Zhai (since 10-2018) CSC 1.0 (supervisor: Telea, Yu)

147 Bernoulli Institute Annual Report

Postdocs C. Ji (since 10-2018) RUG 1.0 (supervisor: Roerdink) L. Yu RUG 0.9 (supervisor: Roerdink, Telea) V. Soancatl Aguilar (since 8-2018) SNN Proeftuin 0.2 (supervisor: Roerdink) A. Sobiecki ZiuZ 0.2 (supervisor: Telea)

Guests X. Ai, China M. Barton,ˇ BCAM, Spain B. Benato, Brasil M. Streit, Johannes Kepler University, Austria Z. Wu, Visiting assistant professor (since 9/2018), Hangzhou Dianzi University, China

148 Bernoulli Institute Annual Report

10.1 Research Program

The research group Scientific Visualization and Computer Graphics carries out research in the area of scientific visualization, information visualization, visual analytics, shape processing, geometric modelling, and computer graphics. With respect to applications, the research concentrates on fundamental and applied problems from the life sciences (in particular medical imaging and bioinformatics), astronomy, large-scale software engineering, and business intelligence.

Large data visualization Visualization of large data sets requires advanced techniques in image processing and seg- mentation, hierarchical data management, and data reduction. Both the increasing size, high dimensionality, and complexity of these data ask for new techniques for interactive visual- ization, so that the speed of the data processing stage matches that of the visualization step. We address this demand by developing efficient algorithms and/or by mapping the involved computations to programmable Graphics Processing Units (GPUs), which are capable of out- performing CPUs for certain compute-intensive applications. We develop scalable techniques for e-Visualization of big data, with a particular focus on time-dependent data, graphs, and networks.

Multiscale shape processing Recent 3D data acquisition and segmentation techniques have made it possible to gather large collections of complex 3D shapes. These come in a variety of formats, such as point clouds, range data, or volumetric densely-sampled fields. Shape processing targets the extraction of high-level information from such datasets, aiming at an easier and better understanding of the structure, topology, and geometry of 3D shapes embedded in the data. We address this using 2D and 3D medial descriptors, or skeletons, which jointly capture shape structure and properties in a compact, multiscale, fashion. We have continued our work on designing fast, scalable, easy to use, and exact 3D skeletonization algorithms, obtaining results that surpass the state-of-the-art in the field. We next apply such methods to a wide range of problems, including 2D and 3D image and shape segmentation, image compression, and 3D shape manipulation.

Multidimensional data exploration Data collections having a large number of attributes recorded per data point, also known as multidimensional datasets, are increasingly present and important in many application fields. We support multidimensional data exploration by using dimensionality-reduction projections, which generate scatterplot-like views of tens of thousands of observations having tens of dimensions. To ease the interpretation of projections, we develop new explanatory techniques that enable end users to interactively retrieve high-dimensional information such as identities

149 Bernoulli Institute Annual Report and ranges of dimensions from the compact, simplified, scatterplot views. We apply our explanatory methods to problems from machine learning and deep learning (classifier design), software understanding, and astrophysics.

Geometric modelling Being able to represent a smooth shape is crucial in modelling. To this end, spline and subdivision techniques are of special importance. We develop new techniques which include locally-refinable spline and subdivision techniques relying on hierarchical spaces, and also con- version methods between various smooth representations such as CAD models and subdivision surfaces.

Vector graphics Vector graphics, as opposed to raster graphics, represent scalable, resolution independent images via certain primitives, such as curves and meshes. We focus on the gradient mesh tool, available e.g. in Adobe Illustrator, and work on extensions supporting local refinement and multi-sided polygons.

Visual Story Telling of Big Imaging Data The main research question in this project is how to develop IT support for diagnostic and decision-making processes based on large and complex imaging data. The approach is based on developing novel graphics, visualization, and interaction methods for the exploration of imaging data. Visual storytelling is an innovative approach for visual presentation and communication that is especially important in situations where the data analyst is not the same person as the decision-maker, and information needs to be exchanged in an intuitive and easy- to-remember way. Diagnostics in radiology will be the primary use case to test the developed approaches. This research is a collaboration with the Netherlands eScience center (NLeSc), the University Medical Center Groningen and the Center for Information Technology (CIT) of the University of Groningen.

150 Bernoulli Institute Annual Report

Figure 5: Left: Classical curve-rendering of 3D DT-MRI tractography dataset, color-coded by local tract direction. Right: Simplified view of the same dataset obtained using our anisotropy- aware bundling method. Fiber sheets, respectively tubular fiber bundles, are now easier to visually separate from each other and explore.

Applications In medical visualization the group studies the analysis and visualization of data obtained by structural and functional imaging techniques such as fMRI, DTI, PET, or EEG. The detection process is complex, requiring image processing to obtain high quality images, mathematical and statistical analysis for quantitative characterization of significant effects, and visualization for interpretation of the results. Work in this area has focused on a the detection and visualization of dynamic EEG coherence networks. The group participates in a research effort on problems from astronomy. Astronomical data sets are growing to enormous sizes. To explore these data sets effectively, new and scalable tools must be developed that can cope with the sheer data volume which has entered the tens of terabytes regime. In this research the focus is on feature extraction, interactive visualization and visual analytics techniques for high-dimensional data. New work focuses on the display of large astronomical data sets in Dome Theaters. We also explore new innovative interactive methods to depict multidimensional abstract data, with two application areas. First, we analyze how multidimensional projections can be used to

151 Bernoulli Institute Annual Report

Figure 6: Left: Overview of the visual storytelling tool with a main window for 2D and 3D data exploration, one widget for changing the settings, and one widget for provenance graph interaction. Right: A provenance graph with multiple branches, thumbnails of the main window in the end-nodes, and image preview and details on-demand (the highlighted end-node). get insight into the operation of machine learning methods such as traditional classifiers and deep neural networks. This approach effectively opens the so-called ‘black box’ of classifier design by allowing users to discover how, where, when, and why such classifiers learn (and operate) effectively. We next complement and extend such techniques with interaction to support visual active learning, a semi-supervised training paradigm where the user collaborates with the machine learning algorithm to reduce the effort needed to obtain a trained engine delivering high accuracy. We apply visual active learning to problems of image classification in biology, in partnership with the University of Campinas, Brazil. We also study the real-time visualization of relational, time-dependent, and multivariate data collected from performance measurement of air traffic controllers in real-life and remote (virtual reality) tower operations, with the aim of understanding the factors which contribute to work stress. This research is done together with the Universities La Sapienza (Rome, Italy) and Toulouse, France.

10.2 Overview of scientific results

Medical data visualization Work on data-driven visualization of multichannel EEG coherence networks was extended to dynamical networks. A new method based on community structure analysis was used to preserve spatial relationships between regions and allow an analysis of the functional

152 Bernoulli Institute Annual Report

Figure 7: To address the issue of local refinement, which is not possible in the available implementations of gradient meshes in Adobe Illustrator, CorelDRAW or Inkscape, we investi- gated an exact patch splitting scheme. In this scheme, any patch can be split horizontally or vertically into smaller patches, allowing for local details without excessive refinement. connectivity within and between brain regions. A method was developed for comparing EEG coherence networks using the earth mover’s distance (EMD) between the distributions of functional units. Earlier work on graph and trail bundling was extended to handle, for the first time, three- dimensional trail sets. A new method was designed to simplify 3D brain tractography datasets obtained from DT-MRI scans. The method takes into account the data anisotropy to separate and also emphasize surface-like regions and tubular regions formed by such tracts (fibers), thereby providing a simplified, easier to visually explore, view of complex tractography datasets. See Figure 5.

Visual Story Telling of Big Imaging Data A method was introduced for the integration of visual storytelling techniques together with provenance data in the analytic systems used in medicine. This can improve the interaction with provenance data displayed in a graph in order to facilitate authoring and the creation process of visual data stories. See Figure 6.

Gradient meshes The gradient mesh primitive is based on a regular rectangular array of bi-cubic patches connected in a smooth way. The user can specify the positions and colours of the vertices of the mesh, as well as so-called colour gradient handles that guide the propagation of colour away from the vertices. While this primitive can be used to produce realistic (and scalable) images, it suffers from several limitations. One such limitation is the rigid topological structure

153 Bernoulli Institute Annual Report and another is the lack of support for local refinement. We have addressed these limitations by introducing a special splitting scheme into the gradient mesh primitive; see Figure 7.

10.3 Research subjects

L. Amabili: Visual storytelling. P. Barendrecht: Vector graphics. X. Chen: Large shape database visualization. M. Espadoto: Visual analytics for deep learning engineering. G.J. Hettinga: Polygonal models and image vectorisation. C. Ji: Visualization of brain connectomics data. Y. Kim: Visual analytics of big data. J. Kosinka: Geometric modelling; computer graphics. L. Pagliosa: Visual analytics for dynamic time series. M. C. Rodriguez: Visualization of classifier decision boundaries. J.B.T.M. Roerdink: Scientific visualization; morphological and wavelet-based multidimen- sional data processing; neuroimaging; bioinformatics. V. Soancatl Aguilar: Visualization in dome theaters. A. Sobiecki: Visual analytics for sensor data. A.C. Telea: Information visualization; software visualization; dimensionality reduction; mul- tiscale shape processing. E. Vernier: Visualization of multidimensional temporal and hierarchical data. J. Wang: Image skeletonisation. L. Yu: Interaction for visual storytelling. X. Zhai: Interactive visual analytics for understanding high-dimensional data.

10.4 Publications

Dissertations

Matthew van der Zwan. Visual Analytics of Multidimensional Time-dependent Trails. • October 5, 2018. A. Telea, M. Wilkinson Chengtao Ji. Visualization and Exploration of Multichannel EEG Coherence Networks. • October 15, 2018. Promotor: J. B. T. M. Roerdink, N. M. Maurits. Venustiano Soancatl Aguilar. Visual Analysis and Quantitative Assessment of Hu- • man Movement. March 19, 2018. Promotor: J. B. T. M. Roerdink, N. M. Maurits, copromotor: C. J. C. Lamoth.

154 Bernoulli Institute Annual Report

Edited publications

A. Telea, T. Theoharis. Proceedings of Eurographics Workshop on 3D Object Retrieval • (3DOR), Eurographics, 2018.

T. Ritschel, A. Telea. Proceedings of Eurographics Tutorials 2018 (EGTut), Eurograph- • ics, 2018.

A. Telea, A. Kerren, J. Braz. Proceedings of the 13th International Joint Conference on • Computer Vision, Imaging and Computer Graphics Theory and Applications (IVAPP), volume 3, SCITEPRESS, 2018.

Articles in scientific journals

P.J. Barendrecht, M. Barton,ˇ J. Kosinka, Efficient quadrature rules for subdivision • surfaces in isogeometric analysis, Computer Methods in Applied Mechanics and Engi- neering, 340, 2018, 1–23. P.J. Barendrecht, M. Luinstra, J. Hogervorst, J. Kosinka, Locally refinable gradient • meshes supporting branching and sharp colour transitions, The Visual Computer, 34(6– 8), 2018, 949–960.

M. Bizzarri, M. Lavi´ cka,ˇ J. Vrsek,ˇ J. Kosinka, A direct and local method for computing • polynomial Pythagorean-normal patches with global G1 continuity, Computer-Aided Design, 102, 2018, 44–51. J.P. Duro Reis, J. Kosinka, Injective hierarchical free-form deformations using THB- • splines, Computer-Aided Design, 100, 2018, 30–38. G.J. Hettinga, J. Kosinka, Multisided generalisations of Gregory patches, Computer • Aided Geometric Design, 62, 2018, 166–180. T.W. Verstraaten, J. Kosinka, Local and Hierarchical Refinement for Subdivision Gradi- • ent Meshes, Computer Graphics Forum, 37(7), 2018, 373–383. Cai, X., Efstathiou, K., Xie, X., Wu, Y., Shi, Y., and Yu, L. A study of the effect of • doughnut chart parameters on proportion estimation accuracy. Computer Graphics Forum 37, 6 (9 2018), 300–312.

Garcia, R., Telea, A., da Silva, B., Torresen, J., and Dihl Comba, J. A task-and-technique • cen- tered survey on visual analytics for deep learning model engineering. Computers & graphics- Uk 77 (12 2018), 30–49.

155 Bernoulli Institute Annual Report

Gronde, van de, J., and Hesselink, W. Conditionally complete sponges: New results on • generalized lattices. Indagationes Mathematicae-New series (11 2018).

Hurter, C., Puechmorel, S., Nicol, F., and Telea, A. Functional decomposition for • bundled simplification of trail sets. IEEE Transactions on Visualization and Computer Graphics 24, 1 (1 2018), 500–510.

Ji, C., Maurits, N. M., and Roerdink, J. B. T. M. Data-driven visualization of multichan- • nel EEG coherence networks based on community structure analysis. Applied Network Science 3, 41 (9 2018).

Matute, J., Telea, A., and Linsen, L. Skeleton-based scagnostics. IEEE Transactions on • Visualization and Computer Graphics 24, 1 (1 2018), 542–552.

Rauber, P., Falcao, A., and Telea, A. Projections as visual aids for classification system • design. Information visualization 17, 4 (10 2018), 282–305.

Soancatl Aguilar, V., Lamoth, C., Maurits, N. M., and Roerdink, J. B. T. M. Assessing • dynamic postural control during exergaming in older adults: A probabilistic approach. Gait & Posture 60 (2 2018), 235–240.

Soancatl Aguilar, V., Martinez Manzanera, O., Sival, D., Maurits, N. M., and Roerdink, • J. B. T. M. Distinguishing patients with a coordination disorder from healthy controls using local features of movement trajectories during the finger-to-nose test. IEEE Trans. Biomedical Engineering (2018).

Soancatl Aguilar, V., van de Gronde, J., Lamoth, C., Maurits, N. M., and Roerdink, J. B. • T. M. Assessing dynamic balance performance during exergaming based on speed and curvature of body movements. IEEE Transactions on Neural Systems and Rehabilitation Engineering (1 2018), 171–180.

Staron, M., Sahraoui, H., and Telea, A. Special section on visual analytics in software • engineering. Information and Software Technology 98 (6 2018), 117–117.

Telea, A., and Kerren, A. Selected papers from IVAPP 2018. AHF-Information 9, 7 (7 • 2018).

Yang, P., Dong, F., Codreanu, V., Williams, D., Roerdink, J. B. T. M., Liu, B., Anvari- • Moghaddam, A., and Min, G. Improving utility of GPU in accelerating industrial applications with user- centered automatic code translation. Ieee transactions on indus- trial informatics 14, 4 (4 2018), 1347–1360.

156 Bernoulli Institute Annual Report

Articles in conference proceedings

L. Amabili, J. Kosinka, M.A.J. van Meersbergen, P.M.A. van Ooijen, J.B.T.M. Roerdink, • P. Svetachov, L. Yu, Improving Provenance Data Interaction for Visual Storytelling in Medical Imaging Data Exploration, EuroVis 2018 — Short Papers, The Eurographics Association, 43–47.

J. Bakker, P.J. Barendrecht, J. Kosinka, Smooth Blended Subdivision Shading, Pro- • ceedings of Eurographics 2018 — Short Papers, The Eurographics Association, 2018, 37–40.

G.J. Hettinga, P.J. Barendrecht, J. Kosinka, A Comparison of GPU Tessellation Strate- • gies for Multisided Patches, Proceedings of Eurographics 2018 — Short Papers, The Eurographics Association, 2018, 45–48.

Ji, C., Maurits, N. M., and Roerdink, J. B. T. M., Visual analysis of evolution of eeg • coherence networks employing temporal multidimensional scaling. In Eurographics Workshop on Visual Com- puting for Biology and Medicine (2018), Eurographics (European Association for Computer Graphics).

Kim, Y., Telea, A., and Trager, S. High-dimensional astronomical data using dimension • reduction. XXX Canary Islands Winter School of Astrophysics : Big Data in Astronomy ; Conference date: 04-11-2018 Through 10-11-2018.

F. C. M. Rodrigues, R. Hirata Jr., A. Telea. Image-based Visualization of Classifier • Decision Boundaries . Proc. SIBGRAPI, 2018, 132-140

E. F. Vernier, J. Comba, A. Telea. A Stable Greedy Insertion Treemap Algorithm for • Software Evolution Visualization. Proc. SIBGRAPI 2018, 221-239

B. Benato, A. Telea, A. Falcao. Semi-Supervised Learning with Interactive Label • Propagation guided by Feature Space Projections. Proc. SIBGRAPI 2018, 67-85

D. Coimbra, T. A. T. Neves, A. Telea, F. V. Paulovich. The Shape of the Game. Proc. • SIBGRAPI 2018, 328-336

A. Telea. Image-Based Graph Visualization: Advances and Challenges A. Telea. Proc. • Graph Drawing 2018, Springer, 1-23

157 Bernoulli Institute Annual Report

10.5 External funding and collaboration

External funding

MOTO – The Remote Tower. Agency: SESAR-H2020 (European Comission). Funding: • one three-year PhD position (2017-2020) jointly supervised by RUG and the Univ. of Toulouse, France.

NWO-NLeSC, project “Visual Storytelling of Big Imaging Data”, jointly with the • University Medical Center Groningen and the Donald Smits Center for Information Technology (CIT). Program ”Disruptive Technologies (DTEC)” of the Netherlands eScience Center (NLeSC), NWO, and Dutch Digital Delta. Period: 2017-2020. Funding: 260 kEuro.

Visualization of Multivariate Hierarchical Time-dependent Data. Funding: CNPq Brazil. • One four-year PhD position (2017-2021) jointly supervised by RUG and the Federal Univ. of Rio Grande do Sul, Brazil.

Target proeftuin: Mining Big Data, funded by SNN (co-applicant). Total budget 400k. •

External collaboration

Roerdink participates in the Groningen Neuroimaging Center of the research school BCN (Behavioural, Cognitive and Neurosciences), and collaborates on visualization problems related to neuroimaging with the Dep. of Neurology (prof. dr. Maurits, prof. dr. K.L. Leenders). He collaborates with the Kapteyn Astronomical Institute of the University of Groningen on visualization of astronomical data (prof. J. van der Hulst, prof. E. Valentijn), and with the Johannes Kepler University Linz, Austria (prof. M. Streit) on visual storytelling.

Telea collaborates with the Institute of Mathematics and Statistics, Univ. of Sao˜ Paulo, Brazil (profs. L. G. Nonato, R. Hirata Jr, N. Hirata) on visual analytics for machine learning. He also collaborates with the Institute of Informatics, Federal Univ. of Rio Grande do Sul, Brazil (prof. J. Comba) in a four-year joint-supervision PhD project funded by CNPq (Brazil), and with the Institute of Informatics, Univ. of Campinas, Brazil (prof. A. Falcao) in a five-year thematic project on visual analytics for machine learning. He also collaborates on shape processing with r. A. Jalba (Eindhoven University of Technology, the Netherlands) and dr. J. Kustra (Philips Research). In information visualization applications for trail analysis, he collaborates with prof. C. Hurter (ENAC/Univ. of Toulouse, France).

158 Bernoulli Institute Annual Report

Kosinka collaborates with the Institute of applied Geometry at Johannes Kepler University (Linz, Austria; prof. B. Juttler)¨ on hierarchical schemes for spline and subdivision surfaces. Further, he collaborates with the Faculty of Applied Sciences at the University of West Bohemia (dr. M. Lavi´ cka)ˇ on various curve and surface interpolation techniques, and with BCAM, Bilbao, Spain (dr. Barton)ˇ on selected topics in geometric modelling. Recently, a new collaboration has started with UMCG (dr. P. van Ooijen) on bone and fracture processing.

10.6 Further information

Roerdink is member of the Graduate School for Behavioral and Cognitive Neurosciences (BCN Groningen) and the Advanced School for Computing and Imaging (ASCI). He is director of the Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence, co-chair of the center for Data Science and Systems Complexity (DSSC) of the Faculty of Science and Engineering. In 2018, he was BI-representative for Informatics Europe, member of the international steering committee on Visual Computing in Biology and Medicine, member of the international steering committee on Biological Data Visualization, member of IPN (Informatics Research Platform Netherlands), and member of the Centre for Medical Imaging North-East Netherlands (CMI-NEN). He is member of ACM and the Eurographics Association, and Senior Member of the IEEE. He was reviewer for a number of international journals. He was member of the PhD reading committee of N. Gravel (University of Groningen) and Theodore´ Chabardes` (Centre de Morphologie Mathematique,´ Fontainebleau).

Telea was on several conference program committees, notably EuroVis, IEEE Visualization, Graph Drawing, VAST, and EuroVA. He is a member of the national research school ASCI, and of the Board of the University of Groningen Graduate School of Science (GSS), and associate editor of the Pattern Recognition Letters journal. In 2018 he held several keynote lectures on data visualization, most notably at the Graph Drawing 2018 symposium, ICT.OPEN, the 3TU Summer School ”Big Data on the Run”, and an invited winter lecture on visualization at the Uppsala University, Sweden.

Kosinka was a member of the Program Committees of the Eurographics Symposium on Geometry Processing (SGP), Geometric Modeling and Processing (GMP), and the Symposium on Solid and Physical Modeling (SPM). And he was a Poster Co-Chair at Eurographics 2018. He is a member of ASCI and the Eurographics Association.

159 Bernoulli Institute Annual Report

160 Bernoulli Institute Annual Report

11 Software Engineering

Group leader: Prof.dr.ir. P. Avgeriou

Tenured staff (BI members) source fte Prof.dr.ir. P. Avgeriou RuG 1.0

Tenure track source fte Dr. M. Lungu RuG 1.0 Dr. V. Andrikopoulos RuG 1.0

Non-tenured staff (BI members) Dr. Ap. Ampatzoglou RuG 0.0

Tenured staff (other) Dr. R. Smedinga RuG 0.3

Honorary professors Prof.dr. T. van der Storm RuG and CWI 0.2 Prof.dr.ir. M. Stal Siemens 0.0

PhD students H. F. Cadavid Rengifo Ubbo Emmius & RuG 1.0 (supervisor: Andrikopoulos/Avgeriou) S. Charalampidou ITEA2 1.0 (supervisor: Avgeriou) G. Digkas Ubbo Emmius & Univ of Mace- 1.0 donia (supervisor: Avgeriou/Lungu) D. Feitosa NUFFIC-CAPES 1.0 (supervisor: Avgeriou) S. Mahdavi-Hezavehi Ubbo Emmius & Linnaeus Univ 1.0 (supervisor: Avgeriou) C. Manteuffel ITEA2 1.0 (supervisor: Avgeriou) A. Reuter RuG 1.0 (supervisor: Andrikopoulos/Avgeriou)

161 Bernoulli Institute Annual Report

D. S. Sas H2020 SDK4ED 1.0 (supervisors: Avgeriou) J. Tan NDSC 1.0 (supervisor: Andrikopoulos, Avgeriou, Lungu) C. Yang Ubbo Emmius & Wuhan Univ 1.0 (supervisor: Avgeriou)

Other PhD students Ar. Ampatzoglou University of external Macedonia (supervisor: Avgeriou) E. M. Arvanitou University of external Macedonia (supervisor: Avgeriou) T. Theunissen HAN external (supervisor: Avgeriou) J. S. van der Ven Independent external Consultant (supervisor: Bosch/Avgeriou) T. Martensson Saab external Aerospace (supervisor: Bosch) A. Hoffman Siemens external (supervisor: Stal)

Guests Slinger Jansen, Utrecht University, the Netherlands Georgia Kapitsaki, University of Cyprus, Cyprus Davide Taibi, Tampere University of Technology, Finland Olaf Zimmermann, HSR Hochschule fur¨ Technik Rapperswil (HSR FHO), Switzerland Danny Weyns, Linnaeus University, Sweden

162 Bernoulli Institute Annual Report

11.1 Research Program

The Software Engineering research program is concerned with theoretical and practical aspects of engineering software and software-intensive systems. The program focuses on one particular field of Software Engineering: Software Architecture. Software Architecture is one of the key disciplines that can help to deal with the hard but also interesting challenges that we are currently facing: increasing integration of Systems and Software Engineering; focus on the end user and the offered added value; increasing demand on software dependability and other critical qualities; dealing with rapid, accelerating change; continuous distribution, mobility, interoperability and globalization; emergence of ultra-large systems (systems of systems); demand for reusability and legacy integration; proliferation of data- and computation-intensive applications; the trend of autonomous or self-managing software; the combinations of biology and computing. The group aims to contribute in architecting and designing industrial software-intensive systems that meet quality standards by carrying out joint research projects with universities, research institutes, and industrial partners, thus combining academic know-how with industrial practice. The group focuses on three major application domains: Embedded Systems (e.g. Healthcare systems), Enterprise Applications and Internet of Things. The group deals with a number of research topics, which are further elaborated in the following paragraphs.

Architectural Knowledge. There is a growing awareness of the importance of Architectural Knowledge (AK) in the software architecture community. Starting with the Griffin project (2005), we have investigated what AK entails, how this can be captured in architectural documentation, the relationship between architectural analysis and AK, and architectural decisions. Documenting architecture decisions has significant benefits for system design and evolution, but is rarely practiced in industry. Therefore, we have developed a framework (published in 2011) that captures all relevant concerns for documenting architecture decisions. In the context of an industry-research collaboration with ABB (2013), we developed a tool for documenting architecture decisions called Decision Architect. The tool has been extended in recent years (2013-2016) and has been released under an open source license. In 2018, we started looking into documentation in Continuous Software Development (CSD) that embraces lean, agile and devops processes. Documentation in CSD is often verbal (like in stand-ups), informal (whiteboard sketches) of completely left out, which leads to knowledge evaporation. We investigate what documentation is required before, during or after an iteration. The result of this research will be a lightweight architectural framework for CSD.

Quality Metrics and Analytics.

163 Bernoulli Institute Annual Report

The object-oriented paradigm is the dominant way in developing software systems at the moment. A basic argument that software developers pose in favor of object-orientation is that it is closer to the way that human brains think, i.e. in terms of objects and actions. However, in order for object-oriented software development to preserve its main advantage, the developed software should reflect a number of quality traits, such as modularity and understandability. The group’s focus in terms of object-oriented design is on methods and techniques that guarantee the internal and external product quality. Our work until now has been focused on GoF design patterns, code refactoring, and quality assessment. In this context, we have: a) validated existing metrics and introduced new metrics b) investigated how GoF design patterns are related to certain quality attributes (e.g., energy efficiency), and what parameters may influence this relation (e.g., pattern grime) c) investigated refactoring techniques and how to prioritize refactorings, and d) developed multiple tools to support quality assessment and refactoring activities.

Technical Debt. Technical debt, which refers to immature software artifacts that fail to meet the required level of quality, has recently attracted increasing attention from both academia and industry in the software engineering field. To date, little work has been done on technical debt at the architecture level, the so-called Architectural Technical Debt (ATD). In the short term, ATD may be incurred to fulfill specific business advantages; but, in the long term, ATD can to a great extent reduce the maintainability and evolvability of a software system. Our group focuses on ATD management in terms of achieving a balance between value and cost of architectural debt. Until now, we have proposed a conceptual model of architectural debt and an ATD management process applying this ATD conceptual model in order to facilitate the decision-making in a value-oriented perspective of architecting. Our current work is focusing on identifying, measuring, and documenting architectural debt as well as economic theories that could be applied to technical debt management and measurement. We have validated that two system-wide modularity metrics can indicate ATD. Another research thread related to ATD is the ranking and prioritization of architectural smells through the information extracted from previous versions of the system by tracking the smells instances themselves. Finally, we are examining the impact of insufficient documentation of the increase of ATD in industrial case studies and investigating the evolution of technical debt in open source software ecosystems.

Architecting Self-Adapting Systems. Self-adaptive systems are resilient and flexible systems capable of autonomously adapting themselves. Due to the continuous evolution of software-intensive systems, a self-adaptive sys- tem monitors itself in order to change its behavior to deal with uncertain operating conditions such as unpredicted system faults, changing stakeholder needs, and changing environment

164 Bernoulli Institute Annual Report and system characteristics. In this context we study a systematic approach for self-adaptation of multiple concerns at runtime. Until now we have systematically reviewed the literature and investigated existing methods. In addition, we have also studied and classified the main uncertainties in the self-adaptive system domain to obtain a better understanding of the prob- lem domain. We are currently focusing on two different studies: 1) developing models that capture the required knowledge of quality concerns and investigating how these models can be employed at runtime to support tradeoff analysis and conflict resolution; and 2) proposing a cost benefit analysis method to include the adaptation cost in runtime decision in order to improve its outcome, and to enable the system to select the optimal configuration for adaptation.

Cloud-based Software Engineering. Cloud computing has become very popular in the last years due to its well-documented benefits to organizations and individuals with respect to transferring capital to operational expenses, potentially unlimited access to computational resources in a self-service manner, and utility-based charging for the use of these resources. In this respect, cloud computing offers a platform for innovative information systems that are partially or completely implemented around cloud offerings. Challenges in this effort arise from the Everything as a Service (*aaS) nature of the platform and the software that is to be developed for it, the volatility of the perceived performance due to its multi-tenant characteristics, the need for cost awareness during operation/execution, and the heterogeneous nature of available offerings for system distribution. These challenges need to be addressed irrespective of the adopted cloud deploy- ment model (public, private, or hybrid), and of whether a system is developed natively on the cloud or migrated to it later. Our research aims to provide the tools and concepts required for engineering software in this context.

Software Language Engineering. Domain-specific languages (DSLs) are little languages, targeted at specific problem domains. DSLs improve communication with domain experts through the use of problem specific notations and abstractions, lead to smaller programs that are easier to maintain, and facilitate reuse of design knowledge across variants in system families. Nevertheless, the construction of DSLs can be a costly endeavour, and use of DSLs is not up to par yet with professional (end-user) development tools. We are investigating techniques to reduce cost of ownership of DSLs by developing generic language workbenches for iterative design and prototyping of software languages. Furthermore, we investigate techniques to facilitate type safe, modular reuse in language engineering. Finally, we’re making the first step towards bringing live programming concepts to the area of DSLs and model-driven engineering; this will allow users of languages to benefit from immediate, run time feedback as they evolve their programs.

165 Bernoulli Institute Annual Report

Architecting Systems of Systems. The term “System of Systems” (SoS) refers to those systems that emerge from the cooperation between a collection of autonomous-yet-interacting constituent systems which deliver func- tionalities that such constituents cannot accomplish on its own. This approach for systems design has been increasingly adopted to address complex missions in important domains like healthcare, transportation, agriculture, and defense, usually with software-intensive and cyber-physical systems as their constituents. However, designing an architecture for a SoS is particularly challenging, when compared with the design of a conventional, centralized, software-intensive system. Unlike the latter, a SoS architecture aims to satisfy Architectural Significant Requirements in highly uncertain environments (caused by the autonomy and independence of its constituents). Furthermore, there are emergent and sometimes undesir- able behaviors that are difficult to anticipate at design time. In this research area, we are investigating novel approaches for the systematic design of architectures of software-intensive systems-of-systems that address these challenges, considering the phases of analysis, synthesis, evaluation, implementation, maintenance, and evolution. Architecting Cloud-enabled Cyber-Physical Systems. Emerging areas like Internet of Things, cloud robotics, vehicular networks, smart factory, and many others reflect a trend towards the development of Cyber-Physical Systems (CPS) which combine devices, sensors and actuators with Cloud Computing. The combination of physical devices with highly scalable Cloud services provides new opportunities for their development and operation, by e.g. offloading complex computation or data analysis tasks to the Cloud in order to offer advanced functionality with low-cost physical devices. Designing the architecture of those systems provides new challenges to software architects of such Cloud- enabled CPS (CeCPS) since the involved technologies range from embedded systems to cloud computing. While embedded systems have limited computing capacity their reaction times are very critical. Cloud computing, on the other hand, offers almost unlimited computing capacity but is subject to network latency. Connecting the two areas requires careful consideration of the system-specific requirements. Another challenge lies in the concerns of multiple stakeholders from different fields which have to be taken into account during the design of the systems software architecture to ensure a secure, productive and maintainable system. To address those challenges, our research is focused on methods and tools that support the decision making in the software architecture design process for CeCPS. Software Ecosystems. Software Ecosystems are groups of software systems that co-evolve together in the same environment. Such an environment can be an organization, an open source collective, or even the entire community formed around a particular programming language. Analyzing ecosystem evolution is critical for the future management of software ecosystems but also

166 Bernoulli Institute Annual Report for improving the evolution of individual systems. Many challenges in software development emerge at the limit between systems and not within a single system itself. These challenges are accentuated in the current context in which inter-system dependencies become dynamic as a result of the raise in adoption of service architectures. One first step towards analyzing, understanding, and controlling ecosystem evolution is developing infrastructure that treats the entire codebase in the ecosystem and the associated artifacts as input and applies data science techniques in this context. As a starting point we are working towards devising techniques and tools for visualizing the evolution of service ecosystems and analyzing the evolution of technical debt in software ecosystems.

Architecting Embedded Systems. Embedded Systems (ESs) are pervasive in modern society. However, the development of these systems is still challenging, as their complexity grows together with the innovations and necessities of modern society. In order to facilitate the development of ESs and guarantee the quality within their design, we performed research on quality assessment. In particular, the research aimed at improving the state-of-the-art on designing critical-embedded systems, a special type of ESs. In Critical Embedded Systems (CES) the assurance of critical quality attributes plays an important role as failures of the system may cause serious damage to the environment, to human lives, to expensive equipment, or non-recoverable financial losses. To this end, we investigated how the use of design patterns is related to multiple critical quality attributes, namely, security, correctness, and performance. In this context, we have found that, when applicable, design patterns can promote critical quality attributes while also supporting the management of noncritical quality attributes (e.g., maintainability). However, phenomena such as pattern grime (i.e., pollution of a pattern instance’s structure) and factors such as the logical and design complexity of pattern instances can affect the extent of the benefits that design patterns could otherwise promote.

11.2 Research subjects

V. Andrikopoulos: cloud native and cloud enabled application engineering, cloud economics, service engineering, systems of systems architecting. P. Avgeriou: software architecture design and evaluation, patterns and pattern languages, architectural knowledge, technical debt, architecture metrics. Ap. Ampatzoglou: object-oriented design, artifact traceability, software quality measurement. Ar. Ampatzoglou: economics, technical debt. E. M. Arvanitou: software quality measurement. H. F. Cadavid Rengifo: architectural design, systems of systems architecting.

167 Bernoulli Institute Annual Report

S. Charalampidou: software metrics, object-oriented refactoring, software artifact traceabil- ity, technical debt management. G. Digkas: architectural design, architectural smells, big software data, technical debt. D. Feitosa: architectural design, object-oriented design and patterns, software quality mea- surement, embedded systems. M. F. Lungu: software ecosystems, software analytics, software evolution. S. Mahdavi-Hezavehi: architectural design, self-adaptive systems. C. Manteuffel: architecture decisions, architecting embedded systems. A. Reuter: cloud-supported cyber-physical systems and system of systems architecture. D. D. Sas: architectural technical debt, architectural smells, software evolution, embedded systems. R. Smedinga: oo-approach, architecture design decision representation. M. Stal: software architecture, architectural patterns. T. van der Storm: software language engineering, programming languages, domain-specific languages. J. Tan: technical debt, software evolution. T. Theunissen: architectural knowledge, continuous software development, agile, lean, de- vops, CI/CD. J.S. van der Ven: software architecture. C. Yang: architectural assumptions, architecture-agility combination.

11.3 Publications

Dissertations

Elvira-Maria Arvanitou. Proposing and Emprically Validating Change Impact • Analysis Metrics. In PhD thesis, University of Groningen. July 2018.

Chen Yang. Architectural assumptions and their management in software devel- • opment. In PhD thesis, University of Groningen. March 2018.

Articles in scientific journals

Aarssen, Rodin; Vinju, Jurgen; van der Storm, Tijs. Concrete Syntax with Black Box • Parsers. In The Art, Science, and Engineering of Programming, 2019, Vol. 3, Issue 3, Article 15.

168 Bernoulli Institute Annual Report

Andrikopoulos, V., Lazovik, A., Kollenstart, M., Langius, E., & Harmsma, E.. Adaptive • Provisioning of Heterogeneous Cloud Resources for Big Data Processing.. In Big Data and Cognitive Computing, 2(3), 2018..

Feitosa, D., Ampatzoglou, A., Avgeriou, P., & Nakagawa, E.Y.. Correlating Pattern • Grime and Quality Attributes.. In IEEE Access, 6, 23065-23078, 2018..

Gomez´ Saez,´ S., Andrikopoulos, V., Bitsaki, M., Leymann, F., & van Hoorn, A.. • Utility-Based Decision Making for Migrating Cloud-Based Applications.. In ACM Transactions on Internet Technology, 18(2, SI), [22]..

Kurs, J., Vrany, J., Ghafari, M., Lungu, M., & Nierstrasz, O.. Efficient parsing with • parser combinators.. In Science of computer programming, 161, 57-88, 2018..

Manteuffel, C., Avgeriou, P., & Hamberg, R.. An exploratory case study on reusing • architecture decisions in software-intensive system projects.. In Journal of Systems and Software, 144, 60-83, 2018..

Shepherd, D.C., & Avgeriou, P.. Stories from the Front.. In Journal of Systems and • Software, 146, A1-A1, 2018..

Yang, C., Liang, P., & Avgeriou, P.. Assumptions and their management in software • development: A systematic mapping study.. In Information and Software Technology, 94, 82-110, 2018..

Yang, C., Liang, P., & Avgeriou, P.. Evaluation of a process for architectural assump- • tion management in software development.. In Science of computer programming, 168, 38-70, 2018..

Articles in International Conferences

Amanatidis, Theodoros; Mittas, Nikolaos; Chatzigeorgiou, Alexander; Ampatzoglou, • Apostolos; Angelis, Lefteris. The developer’s dilemma: Factors affecting the Deci- sion to Repay Code Debt. In In: Proceedings of the 2018 International Conference on Technical Debt (TechDebt ’18), pp. , New York, NY, USA: ACM Press, 2018..

Ampatzoglou, A., Bibi, S., Chatzigeorgiou, A., Avgeriou, P., & Stamelos, I.. Reusabil- • ity Index: A Measure for Assessing Software Assets Reusability.. In In: Interna- tional Conference on Software Reuse ICSR 2018: New Opportunities for Software Reuse, pp. 43-58, Springer, 2018..

169 Bernoulli Institute Annual Report

Ampatzoglou, A., Michailidis, A., Sarikyriakidis, C., Ampatzoglou, A., Chatzigeorgiou, • A., & Avgeriou, P.. A Framework for Managing Interest in Technical Debt: An Industrial Validation.. In In: Proceedings of the 2018 International Conference on Technical Debt (TechDebt ’18), pp. 115-124, New York, NY, USA: ACM Press, 2018..

Back, T., & Andrikopoulos, V.. Using a Microbenchmark to Compare Function as a • Service Solutions.. In In: Service-Oriented and Cloud Computing: 7th IFIP WG 2.14 European Conference, ESOCC 2018, Como, Italy, September 12-14, 2018, Proceedings, pp. 146-160, Springer, 2018..

Charalampidou, S., Ampatzoglou, A., Chatzigeorgiou, A., & Tsiridis, N.. Integrat- • ing Requirement Specifications and Source Code Traceability within the IDE to Prevent Documentation Debt.. In In: 44th Conference on Software Engineering and Advanced Applications (SEAA), pp.421-428, IEEE, 2018..

Charalampidou, S., Arvanitou, E-M., Ampatzoglou, A., Chatzigeorgiou, A., Avgeriou, • P., & Stamelos, I.. Structural Quality Metrics as Indicators of the Long Method Bad Smell: An Empirical Study.. In In: 44th Conference on Software Engineering and Advanced Applications (SEAA), IEEE computer society, 2018..

Coulon, Fabien; Degueule, Thomas; van Der Storm, Tijs; Combemale, Benoit;. Shape- • diverse DSLs: languages without borders (vision pape). In Proceedings of the 11th ACM SIGPLAN International Conference on Software Language Engineering (SLE’18) pp. 215–219. Distinguished Vision Paper Award..

Digkas, G., Lungu, M., Avgeriou, P., Chatzigeorgiou, A., & Ampatzoglou, A.. How • do developers fix issues and pay back technical debt in the Apache ecosystem? In 2018 IEEE 25th International Conference on Software Analysis, Evolution and Reengineering (SANER) (pp.. In In: 2018 IEEE 25th International Conference on Software Analysis, Evolution and Reengineering (SANER), pp.153-163, IEEE, 2018..

Feitosa, D., Ampatzoglou, A., Avgeriou, P., Affonso, F.J., Andrade, H., Felizardo, • K.R., & Nakagawa, E.Y.. Design Approaches for Critical Embedded Systems: A Systematic Mapping Study.. In In: Evaluation of Novel Approaches to Software Engineering: 12th International Conference, ENASE 2017, Porto, Portugal, April 28–29, 2017, Revised Selected Papers, pp.243-274, Springer, 2018..

Inostroza, Pablo; van der Storm, Tijs. JEff: objects for effect. In Proceedings of the • 2018 ACM SIGPLAN International Symposium on New Ideas, New Paradigms, and Reflections on Programming and Software (Onward!) pp. 111–124.

170 Bernoulli Institute Annual Report

Lampropoulos, Alexander; Ampatzoglou, Apostolos; Bibi, Stamatia; Chatzigeorgiou, • Alexander; Stamelos, Ioannis. REACT: A Process for Improving Open-Source Soft- ware Reuse. In In: 11th International Conference on the Quality of Information and Communications Technology (QUATIC’ 18), 2018..

Paschali, Maria Eleni; Bafatakis, Nikolaos; Ampatzoglou, Apostolos; Chatzigeor- • giou, Alexander; Stamelos, Ioannis. Tool-Assisted Game Scenario Representation through Flow Charts. In In: 13th International Conference on the Evaluation of Novel Approaches to Software Engineering (ENASE’18), 2018..

Skiada, Peggy; Ampatzoglou, Apostolos; Arvanitou, Elvira-Maria; Chatzigeorgiou, • Alexander; Stamelos, Ioannis. Exploring the Relationship between Software Mod- ularity and Technical Debt. In In: 44th Conference on Software Engineering and Advanced Applications (SEAA), IEEE computer society, 2018..

Tikhonova, Ulyana; Stoel, Jouke; van Der Storm, Tijs; Degueule, Thomas. Constraint- • based run-time state migration for live modeling. In Proceedings of the 11th ACM SIGPLAN International Conference on Software Language Engineering (SLE’18) pp. 108–120. VERSEN Best Software Engineering Paper in the Netherlands.

Verano Merino, Mauricio; Vinju, Jurgen; van der Storm, Tijs. Bacata:´ a language para- • metric notebook generator (tool demo). In Proceedings of the 11th ACM SIGPLAN International Conference on Software Language Engineering (SLE’18) pp. 210–214.

Articles in Peer-Reviewed International Workshops and Symposia

Andrikopoulos, V.. Engineering Cloud-based Applications: Towards an Applica- • tion Lifecycle.. In In: Advances in Service-Oriented and Cloud Computing: Workshops of ESOCC 2017, Oslo, Norway, September 27-29, 2017, Revised Selected Papers, pp.57-72, Springer, 2018..

Arvanitou, E.M., Ampatzoglou, A., Tzouvalidis, K., Chatzigeorgiou, A., Avgeriou, P., & • Deligiannis, I.. Assessing change proneness at the architecture level: An empirical validation.. In In: 2017 24th Asia-Pacific Software Engineering Conference Workshops, APSECW 2017: Proceedings, pp. 98-105, IEEE, 2018..

Papadopoulos, Lazaros; Marantos, Charalampos; Digkas, Georgios; Ampatzoglou, • Apostolos; Chatzigeorgiou, Alexander; Soudris, Dimitrios. Interrelations between Software Quality Metrics, Performance and Energy Consumption in Embedded Applications. In In: 21st International Workshop on Software and Compilers for Embedded Systems (SCOPES 18), 2018..

171 Bernoulli Institute Annual Report

Verano Merino, M., van der Storm, T.. Language workbench support for block- • based DSLs. In BLOCKS+ Workshop 2018, collocated with SPLASH’18..

Other publications

Smedinga, R., & Biehl, M.(Eds.). 15th SC@RUG 2018 proceedings 2017-2018.. In • Rijksuniversiteit Groningen, 2018..

172 Bernoulli Institute Annual Report

11.4 External funding and collaborations

External funding The group has obtained one new project (PI: Avgeriou): Visual diagnosis for DevOps software development (VISDOM), funded with 1 PhD position and 1 postdoc by ITEA. Within the European consortium, our group focuses specifically on developing methods and tools to measure and visualize quality of software (particularly technical debt). The group has continued work (PI: Avgeriou) on the project Software Development toolKit for Energy optimization and technical Debt elimination (SDK4ED), funded with 1 PhD position from H2020-ICT-2017-1. Within the European consortium, our group focuses on developing a set of software tools for preventing the degradation of run-time qualities and especially energy consumption, while allowing for efficient measuring of the accumulated Technical Debt during the development of new low-energy computing software applications, including embedded systems and IoT products. External collaborations The group worked together with the University of Sao˜ Paulo in Brazil. In the context of the SDK4ED project we work with Airbus, Imperial College London, Tiobe, Maxeler, CERTH, Holisun, CNET, Neurasmus, National Technical University of Athens and University of Macedonia. We have an informal project with the University of Macedonia (Greece), where Elvira Maria Arvanitou and Areti Ampatzoglou are both located and co-supervised. Finally, we have joint PhD projects with Linnaeus University (Sweden) [S. Mahdavi-Hezavehi], Wuhan University (China) [Y. Chen], and the University of Macedonia (Greece) [G. Digkas]. V. Andrikopoulos and A. Reuter are collaborating with Elena Lazovik and Johan van der Geest on behalf of TNO on the subject of architectures for large scale distributed systems. In addition to the above, the group has collaborated with the following foreign universities and organizations in terms of joint publications, co-supervision of theses and research projects: Wuhan University, China (Prof. Liang), Swinburne University, Australia (Dr. Tang), University Rey Juan Carlos, Spain (Dr. Capilla), Limerick University, Ireland (Prof. Fitzgerald), Univer- sity of Vienna, Austria (Prof. Zdun), Katholieke Universiteit Leuven, Belgium (Prof. Holvoet), University of Canterbury, New Zealand (Dr. Galster), Siemens AG, Germany (Prof. Stal), Tampere University of Technology, Finland (Prof. Koskimies, Prof. Taibi), Linnaeus Uni- versity, Sweden (Prof. Weyns and Andersson), Chalmers University of Technology, Sweden (Prof. Bosch and Chaudron), HSR, Switzerland (Prof. Zimmermann), Polytechnic University

173 Bernoulli Institute Annual Report of Catalunya, Spain (Prof. Franch), University of British Columbia, Canada (Prof. Kruchten), University of Helsinki, Finland (Prof. Mannisto), University of South Brittany, France (Prof. Oquendo), University of Pretoria, South Africa (Dr. Solms), Koopman (P. Oosterhoff) and Sustainable Buildings (Dr. T.A. Nguyen and Dr. F. Nizamic), the Netherlands University of Bern (Prof. Nierstrasz), University of Hong Kong (Prof. Oliveira), University of Athens (Mr. Biboudis). The group has ongoing Socrates-Erasmus agreements together with the Tampere Technical University (Finland), Universidad Rey Juan Carlos (Spain), Linnaeus University (Sweden), University of L’Aquila (Italy), University of Piraeus, (Greece), University of Cyprus (Cyprus), Wroclaw University of Technology (Poland), University of Macedonia (Greece), Aristotle University (Greece), National Technical University of Athens (Greece). Tijs van der Storm runs an INRIA Associate Team with Prof. Benoit Combemale from Toulouse.

11.5 Further activities

P. Avgeriou is the Editor-in-Chief of the Journal of Systems and Software and an Associate Editor of IEEE Software. He was co-organizer of two international workshops: the 6th International Workshop on Software Engineering for Systems-of-Systems (SESoS 2016), as well as the Special Session on Software Engineering and Technical Debt (SEaTeD 2018). He also co-organized the second International Software Architecture PhD School (iSAPS), at the Lorentz Center, while he co-chaired the doctoral symposium at the European Conference on Software Architecture (ECSA). He served in numerous Program and Steering Committees of international conferences and in the editorial board of Springer Transactions on Pattern Languages of Programming. He served in the Board of the Research School ‘Institute for Programming and Algorithmics’ (IPA), in the board of the Dutch National Association for Software Engineering (VEReniging Software Engineering Nederland – VERSEN) and the advisory board of the Groningen Engineering Center. He has continued service as a member of the ICT Research Platform Nederland (IPN), the International Software Engineering Research Network (ISERN), as Senior Member of IEEE and as a member of the Royal Institute for Engineers (KIVI). V. Andrikopoulos gave invited talks at

the ICT Open 2018 conference, accompanied by a poster for the VERSEN track. • the RUG/Osaka University collaboration workshop in March 2018. •

174 Bernoulli Institute Annual Report

Together with Paris Avgeriou he participated in the university delegation visiting the Google datacenter at Eemshaven for potential collaboration in February 2018, He was a member of the organization committee of the 4th Dutch national symposium on software engineering (SEN Symposium 2018). He acted as PhD Symposium co-chair at the 7th European Conference on Service-Oriented and Cloud Computing (ESOCC 2018), and as the CloudWays 2018 workshop co-chair, also in the context of ESOCC 2018. He served in one PhD committee for the University of Technology Sydney, and in one PhD defense committee for the University of Groningen. Since 2018 he is a member of VERSEN and IEEE, and of the IFIP Working Group on Service Oriented Systems. T. van der Storm is currently treasurer of the European Association for Programming Languages and Systems (EAPLS), chair of the IFIP TC2 Working Group 2.16 on Language Design, and member of the Association Internationale pour les Technologies Objets (AITO). He served on program committees of international conferences, such as SLE’18, GPCE’18, ECOOP’18, and Scala’18, and was member of reviewing board of the journal on the The Art, Science, and Engineering of Programming.

11.6 Distinctions

Paris Avgeriou was ranked in the top 5 of Software Engineering authors worldwide in the topic of Architecture, in a bibliographic survey published at IEEE Software (vol. 5, 2018). Paris Avgeriou gave a keynote speech at the 18th IEEE International Conference on Software Quality, Reliability, and Security (QRS), Lisbon, Portugal, July 16-20 2018. Paris Avgeriou received a ’best paper award’ for the paper ’Reusability Index: A Measure for Assessing Software Assets Reusability’, at the 17th International Conference on Software Reuse (ICSR), May 21-23, 2018, Madrid, Spain.

175 Bernoulli Institute Annual Report

176 Bernoulli Institute Annual Report

12 Autonomous Perceptive Systems

Group leader: Prof. dr. L.R.B. Schomaker

Tenured staff (BI members) source fte Prof. dr. L.R.B. Schomaker RUG 1.0 Prof. dr. R. Carloni RUG 1.0 Dr. S.M. Netten, van RUG 1.0

Tenure track assistant professors Dr. M. Wiering RUG 1.0

PhD students M. Ameryan RUG 1.0 (supervisor: Schomaker, Wiering) Y. Chen PhD Scholarship 1.0 (supervisor: Schomaker, Wiering) K. Dijkstra NHL 1.0 (supervisor: van de Loosdrecht, Schomaker, Wiering) M.A. Dhali NHL 0.8 (supervisor: Schomaker, Popovic) V. Krishna starting December RUG 1.0 (supervisor Carloni, Focarete) S. Luo PhD Scholarship 1.0 (supervisor: Schomaker, Wiering) A. Mazumder starting May RUG 1.0 (supervisor: Carloni, Zucchelli) E. Okafor PhD Scholarship 1.0 (supervisor: Schomaker, Wiering) P. Pawara PhD Scholarship 1.0 (supervisor: Schomaker, Wiering) V. Raveendranathan (starting April) RUG 1.0 (supervisor: Carloni, Zucchelli) A.E.M. Schmerbauch starting March PhD Scholarship 1.0 (supervisor: Carloni, Zucchelli)

177 Bernoulli Institute Annual Report

H. Wen starting February PhD Scholarship 1.0 (supervisor: Schomaker, Wiering) B.J. Wolf PhD Employed RUG 1.0 (supervisor: Schomaker, van Netten, Andringa) P. Pawara PhD Scholarship 1.0 (supervisor: Schomaker, Wiering)

Postdocs Dr. H. Kasaei (starting August) Fellow RUG 1.0 (supervisor: Schomaker) Dr. P. Pirih RUG 1.0 (supervisor: Schomaker)

Guests S. Rezaee Oshternian, Shiraz University of Technology, Shiraz, Iran

178 Bernoulli Institute Annual Report

12.1 Research Program

The research group Autonomous Perceptive Systems aims at understanding the mechanisms that enable autonomous systems to produce adequate responses when confronted with a complex, time-variant environment. The research addresses three facets: perception, learning and control. The paradigmatic example problems concern classification of images and other sensory patterns, learning of complicated tasks using minimal training information and adaptive control in robotic systems. Major methods are (deep) neural networks, reinforcement learning, adaptive control and explicit biophysical modeling of sensor systems.

Lifelong machine learning Current examples of deep learning application are often impressive, but based on closed data sets. Such data is very clean, heavily curated in terms of segmentation (i.e., the selection of relevant material) and labeling. This is in stark contrast with real-life problems in AI-based robotics and in industrial processes. Here, the amount of labeled material is usually limited and class labels are varying over time. There may be drifts in raw-data properties (sensors, pre-processing methods that are varying over time, etc.). Such problems are not solved, at all. Application domains are in historical document analysis and robotics.

Machine learning for historical document-image analysis The difficult area of digital document-image analysis in historical collections is a fruitful breeding ground for the evaluation and improvement of (deep) machine-learning methods. The group maintains an e-Science cloud service (Monk) for users in the humanities to label and index historical manuscripts. This service represents an observatory for image,label tuples. The problems addressed range from image preprocessing, image-layout analysis, segmentation to (handwritten) text recognition. Other machine-learning tasks are document dating and writer identification. The problems encountered are characterized by a notorious lack of labeled data when a new historical collection in an unknown script style and language is ingested. This means that current deep-learning methods cannot immediately be used and bootstrapping algorithms need to be developed, to reach a critical mass of labeled data.

Robotics The Robotics group develops systems that are intended to physically interact with uncertain dynamic environments and to cooperate with humans. The group’s main focus is the de- velopment of novel actuation systems, which are the key enabling components for motion generation. The work is accomplished by developing unique mechanical designs and intel- ligent control architectures. The main research lines are: 1. Lower-limb prosthetic devices (mechatronic design, modeling, control, experimental validation) and 2. Bio-inspired soft robots/actuators (mechatronic design, modeling, control, experimental validation). Apart from the fundamental scientific problems and the goal of advancing the state of the art, the group

179 Bernoulli Institute Annual Report considers societal relevance very important. This plays a role in the involvement of patients, hospitals, and relevant companies in the biomedical field). Furthermore robotics research requires multisciplinarity and cross-fertilization with other relevant research domains (e.g., chemistry and material science). Robots are also used by the machine-learning researchers in the APS group for experiments that are focused on specific machine-learning problems in object grasping.

Biophysical modelling Artificial mechano-sensory arrays, inspired by the sensory modality of the fish lateral line, are designed, built and tested. These flow sensing arrays consist of all-optical flow sensors, capable of measuring fluid flow ranging from tens of micrometers- to meters per second, and have been tested in increasingly realistic environments (lab, pool, and recently under harsh benthic conditions).

Deep reinforcement learning Reinforcement learning algorithms enable a software agent to learn from its interaction with an environment. The goal is to optimize a policy that obtains the largest sum of rewards in its future. For this, the agent is situated in an environment and observes the environmental state. This state is used to select an action, after which the agent transits to a next state and may receive a reward or punishment. in our research, we have put the main focus on learning to play games. This allows for a controlled environment, simple rules, and huge state spaces. The agent is usually equipped with a (deep) neural network to learn to generalize over the continuous or high-dimensional state space. In 2018, we published 12 papers on RL, 9 of which are on playing games. Furthermore, we published several papers on deep learning and computer vision.

12.2 Overview of scientific results

Monk project of harvesting text-image labels for machine learning In 2018, the threshold of 1 million labeled word images was reached, in over 560 books and manuscripts, ranging from Chinese characters, Arabic script, Hebrew, Egyptian hieroglyphs to Western medieval and contemporary script styles. With this critical mass of data it will be possible to perform large field tests of the deployment of deep-learning methods and contrast them with traditional approaches. A hot topic in deep learning currently is zero-shot learning, where samples are classified from classes of objects that did not exist in the training set. This was tested in the area of medieval handwritten word images. A second study provided excellent results in Chinese character recognition using rule-based attributes as opposed to the (impossible) one-hot encoding for neural-network classification of Chinese with its massive

180 Bernoulli Institute Annual Report number of pattern classes. Also here, zero-shot learning appeared to be possible.

Predictive maintenance Success was obtained in predicting diesel-engine message patterns in a global fleet of exca- vators, sending status message streams via SIM cards to a central office, from all continents. Using long short-term memory (LSTM) recurrent neural networks with a bottleneck layer for dimensionality reduction, high a accuracy was reached in predicting the next event status value per time step. These results will be extended in the direction of cost prediction using Poisson models of maintenance-event costs.

Robotics & prosthetics The robotics group is currently in the process of starting up two large European projects (MyLeg and Magnify), each 1.0 MEuro. The projects are in the area of adaptive-leg prosthetics in humans.

Machine learning in robotics Machine-learning researchers in the APS group studied the effect of using deep learning on robot navigation in the home environment. It was found that the usual enforcement of geometric invariance by data augmentation (image flipping and rotation) and feature-map subsampling may be beneficial in academic image-retrieval benchmarks for deep learning but such an invariance is undesirable in navigating robots: Left vs Right and Up vs Down actually do matter, for a robot learning the visual landmarks in an indoor environment.

Biophysics A variety of neural-network methods (ESN, MLP, ELM, CNN, LSTM) was used for trans- forming the measured flow signals into hydrodynamic imaging, i.e., determining location and shape of submerged moving objects. Thus far, imaging capabilities have been demonstrated under flow conditions with Reynolds numbers up to the order of 105. Possible applications may be found in, e.g., underwater surveying or navigating by autonomous underwater vehicles (AUVs).

Reinforcement learning Two novel reinforcement-learning algorithms were developed and reported: (a) Deep QV- learning (DQV) and (b) Sampled Policy Gradient (SPG). DQV uses temporal-difference learning to train a Value neural network and uses this network for training a second Quality- value network that learns to estimate state-action values. We tested DQV’s update rules with Multilayer Perceptrons as function approximators on two classic RL problems and then extended DQV by means of deep-convolutional neural networks. The SPG method concerns a new offline actor-critic learning algorithm. It performs a sampling of the action space to calculate an approximated policy gradient by using the ’critic’ component in order to evaluate

181 Bernoulli Institute Annual Report the samples. This sampling allows SPG to search the action-Q-value space more globally than deterministic policy gradient (DPG), theoretically enabling it to avoid local optima more effectively.

12.3 Research subjects

L. Schomaker: Lifelong (deep) learning, historical manuscript image understanding, pattern recognition. G. Maillette de Buy Wenniger: Deep learning and computational linguistics for handwriting recognition, machine translation. M. Dhali Image preprocessing, writer identification and document dating in the Dead Sea Scrolls. M. Ameryan Handwriting recognition in multimodal manuscripts with text and drawings (NWO, Naturalis) S. Luo Robotic object grasping using reinforcement learning. B. Sriman: Camera-based text recognition in urban scene images J.-P. van Oosten: Limitations of Markov-model estimation; non-linear cascades in machine learning processes. R. Carloni: Robot prosthetics, systems & control H. Kasaei: Vision-based object grasping A. Mazumder: (PhD MyLeg) V. Raveendranathan: (PhD MyLeg) M. Wiering: Reinforcement learning and deep learning K. Dijkstra: Deep learning for application of drone vision in agriculture P. Pawara: Deep learning for plant classification Y. Chen: (PhD Wiering) H. Wen: (PhD Wiering) E. Okafor: Deep learning for animal classification

182 Bernoulli Institute Annual Report

A. Shantia: Indoor robot vision and navigation S. van Netten: Biophysics and neural-network modeling of lateral-line sensors B. Wolf: Lateral-line based sensing and imaging P. Piri: Hydrodynamic imaging, insect vision, signal and image processing T. Engbersen: Big Data, computer architecture, applications of AI S. Bohte: Spiking-neural network modeling and reinforcement learning

183 Bernoulli Institute Annual Report

12.4 Publications

Articles in scientific journals

M. Aslani, S. Seipel, M. Mesgari, and M. Wiering. Traffic signal optimization through discrete and continuous reinforcement learning with robustness analysis in downtown tehran. Advanced engineering informatics, 38:639–655, 10 2018.

M. Aslani, S. Seipel, and M. Wiering. Continuous residual reinforcement learning for traffic signal control optimization. Canadian Journal of Civil Engineering, pages cjce–2017–0408, 2018.

R. Carloni, V. Lapp, A. Cremonese, J. Belcari, and A. Zucchelli. A variable stiffness joint with electrospun p(vdf-trfe-ctfe) variable stiffness springs. IEEE Robotics and Automation Letters, 3(2):973–978, 4 2018.

K. Dijkstra, J. van de Loosdrecht, L. Schomaker, and M. Wiering. Hyperspectral demosaicking and crosstalk correction using deep learning. Machine Vision and Applications, 30(1), 7 2018.

S. He and L. Schomaker. Open set chinese character recognition using multi-typed attributes. ArXiv, 8 2018. 29 pages, submitted to Pattern Recognition.

U. Moschini, A. Meijster, and M. Wilkinson. A hybrid shared-memory parallel max-tree algorithm for extreme dynamic-range images. IEEE transactions on pattern analysis and machine intelligence, 40(3):513–526, 3 2018.

E. Okafor, L. Schomaker, and M. Wiering. An analysis of rotation matrix and colour constancy data augmentation in classifying images of animals. Journal of Information and Telecommunication, 2(4):465–491, 2018.

P. Pirih, M. Ilic, J. Rudolf, K. Arikawa, D. Stavenga, and G. Belusic. The giant butterfly- moth paysandisia archon has spectrally rich apposition eyes with unique light-dependent photoreceptor dynamics. Journal of Comparative Physiology A, 204(7):639–651, 7 2018.

D. Stavenga, H. Leertouwer, A. Meglic, K. Draslar, M. Wehling, P. Pirih, and G. Belusic. Classical lepidopteran wing scale colouration in the giant butterfly-moth. PeerJ, 6, 4 2018.

R. Unal, S. Behrens, R. Carloni, E. Hekman, S. Stramigioli, and B. Koopman. Conceptual design of a fully passive transfemoral prosthesis to facilitate energy-efficient gait. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 26(12):2360–2366, 12 2018.

184 Bernoulli Institute Annual Report

K. van den Bosch, T. Andringa, W. Post, W. Ruijssenaars, and C. Vlaskamp. The relationship between soundscapes and challenging behavior: A small-scale intervention study in a health- care organization for individuals with severe or profound intellectual disabilities. Building Acoustics, 25(2):123–135, 6 2018.

K. van den Bosch, D. Welch, and T. Andringa. The evolution of soundscape appraisal through enactive cognition. Frontiers in Psychology, 9, 7 2018.

B. Wolf, J. Morton, W. MacPherson, and S. van Netten. Bio-inspired all-optical artificial neuromast for 2d flow sensing. Bioinspiration & biomimetics, 13(2), 3 2018.

Articles in conference proceedings

M. Aslani, M. Mesgari, S. Seipel, and M. Wiering. Developing adaptive traffic signal control by actor-critic and direct exploration methods. Proceedings of the Institution of Civil Engineers - Transport, 2018.

E. Barrett, M. Reiling, S. Mirhassani, R. Meijering, J. Jager, N. Mimmo, F. Callegati, L. Mar- coni, R. Carloni, and S. Stramigioli. Autonomous battery exchange of uavs with a mobile ground base. pages 699–705, 5 2018. 2018 IEEE International Conference on Robotics and Automation (ICRA) ; Conference date: 21-05-2018 Through 25-05-2018.

F. Bidoia, M. Sabatelli, A. Shantia, M. Wiering, and L. Schomaker. A deep convolutional neural network for location recognition and geometry based information. In M. De Marsico, G. Sanniti di Baja, and A. Fred, editors, Proceedings of the 7th International Conference on Pattern Recognition Applications and Methods, pages 27–36. SciTePress, 2018.

S. Chanda, J. Baas, D. Haitink, S. Hamel, D. Stutzmann, and L. Schomaker. Zero-shot learning based approach for medieval word recognition using deep-learned features, pages 345–350. Proceedings of International Conference on Frontiers in Handwriting Recognition, ICFHR. Institute of Electrical and Electronics Engineers Inc., 12 2018.

S. Chanda, E. Okafor, S. Hamel, D. Stutzmann, and L. Schomaker. Deep learning for classification and as tapped-feature generator in medieval word-image recognition. In 13th IAPR International Workshop on Document Analysis Systems (DAS), pages 217–222. IEEE, 6 2018.

S. Knegt, M. Drugan, and M. Wiering. Opponent modelling in the game of tron using reinforcement learning, volume 2 of ICAART 2018 - Proceedings of the 10th International Conference on Agents and Artificial Intelligence. SciTePress, 2018.

185 Bernoulli Institute Annual Report

J. Kormelink, M. Drugan, and M. Wiering. Exploration methods for connectionist Q-learning in bomberman, pages 355–362. ICAART 2018 - Proceedings of the 10th International Conference on Agents and Artificial Intelligence. SciTePress, 2018.

G. Leuenberger and M. Wiering. Actor-critic reinforcement learning with neural networks in continuous games. ICAART 2018 - Proceedings of the 10th International Conference on Agents and Artificial Intelligence. SciTePress, 2018.

R. Niel, J. Krebbers, M. Drugan, and M. Wiering. Hierarchical reinforcement learning for real-time strategy games, volume 2 of ICAART 2018 - Proceedings of the 10th International Conference on Agents and Artificial Intelligence, pages 470–477. SciTePress, 2018.

E. Okafor, G. Berendsen, L. Schomaker, and M. Wiering. Detection and Recognition of Badgers Using Deep Learning, pages 554–563. Lecture Notes in Computer Science book series. Springer International Publishing, Cham, Switzerland, 10 2018.

P. Ozkohen, J. Visser, M. van Otterlo, and M. Wiering. Learning to play donkey kong using neural networks and reinforcement learning, pages 145–160. Communications in Computer and Information Science. Springer Verlag, 2018.

M. Pieters and M. Wiering. Comparison of machine learning techniques for multi-label genre classification, pages 131–145. Communications in Computer and Information Science. Springer International Publishing AG, 2018.

M. Pieters and M. Wiering. Comparing generative adversarial network techniques for image creation and modification. ArXiv, 3 2018. 20 pages, 23 figures.

M. Sabatelli, F. Bidoia, V. Codreanu, and M. Wiering. Learning to evaluate chess positions with deep neural networks and limited lookahead. 1 2018. 7th International Conference on Pattern Recognition Applications and Methods ; Conference date: 16-01-2018 Through 18-01-2018.

J. van de Wolfshaar, M. Wiering, and L. Schomaker. Deep learning policy quantization. In Proceedings of the 10th International Conference on Agents and Artificial Intelligence, pages 122–130. SciTePress, 2018.

P. van Lenthe, S. Verros, E. Hekman, R. Carloni, and H. Koopman. Comparing assistive admittance control algorithms for a trunk supporting exoskeleton. pages 2828–2834, 5 2018. 2018 IEEE International Conference on Robotics and Automation (ICRA) ; Conference date: 21-05-2018 Through 25-05-2018.

186 Bernoulli Institute Annual Report

B. Verheij and M. Wiering. Artificial Intelligence: 29th Benelux Conference, BNAIC 2017, Groningen, The Netherlands, Revised Selected Paper. Communications in Computer and Information Science. Springer, 1 edition, 2018.

A. Weber, M. Ameryan, K. Wolstencroft, L. Stork, M. Heerlien, and L. Schomaker. Towards a Digital Infrastructure for Illustrated Handwritten Archives, pages 155–166. Springer, 5 2018.

Other publications

E. Okafor and L. Schomaker. Integrated dimensionality reduction and sequence prediction using lstm. 2018. ICT.Open 2018, Amersfoort, The Netherlands.

P. van der Meulen, B. Wolf, P. Pirih, and S. van Netten. Performance of neural networks in source localization using artificial lateral line sensor configurations. 3 2018. ICT OPEN 2018: The Interface for Dutch ICT-Research ; Conference date: 19-03-2018 Through 20-03-2018.

S. van Netten, B. Wolf, and W. MacPherson. Sensor element and method for measuring of near-field, large-scale hydrodynamic characteristics, 11 2018. EP3399320 / European Patent.

187 Bernoulli Institute Annual Report

12.5 External funding and collaboration

External funding

Ongoing projects

H2020 ECSEL/EU Mantis / Schomaker (EU grant 662189). • Cyber Physical System based Proactive Collaborative Maintenance The overall concept of MANTIS is to provide a proactive maintenance service plat- form architecture based on Cyber Physical Systems that allows to estimate future performance, to predict and prevent imminent failures and to schedule proactive main- tenance. Physical systems (e.g. industrial machines, vehicles, renewable energy as- sets) and the environment they operate in, are monitored continuously by a broad and diverse range of intelligent sensors, resulting in massive amounts of data that characterise the usage history, operational condition, location, movement and other physical properties of those systems. https://cordis.europa.eu/project/ rcn/198079/factsheet/en http://www.mantis-project.eu/ MyLeg/EU Carloni (EU grant 780871) P.I. - 4.5 MEuro • Smart and intuitive osseointegrated transfemoral prosthesis embodying advanced dy- namic behaviours The project aims at developing an Osseointegrated Implant, enhancing human-prosthesis interaction, perception, and motion capabilities. Implantable myoelectric sensors on tar- geted reinnervated muscles provide intuitive control. The project will develop variable- stiffness actuators & novel composite materials achieving energy efficiency, dependabil- ity, and adaptability to different tasks. http://www.myleg.eu/ H2020 Lakshmi/ van Netten (EU grant 635568) • Sensors for LArge scale HydrodynaMic Imaging of ocean floor The project will develop a new bio-inspired technology to make continuous and cost- effective measurements of the near-field, large-scale hydrodynamic situation, for en- vironmental monitoring in cabled ocean observatories, marine renewable energy and port/harbor security. https://www.lakhsmi.eu Monk project (various grants, Schomaker) • The Monk project is focused on the development of large-scale image and text search functions for massive collections of historical manuscript. The basic paradigm entails lifelong machine learning which is necessary because new data are arriving from archives and libraries in a continuous stream and the text annotations and transcriptions are evolving over time. This umbrella project encompasses several grants obtained since 2005.

188 Bernoulli Institute Annual Report

NWO/Making Sense of Illustrated Handwritten Archives / Schomaker • Many handwritten and illustrated archives contain a wealth of information, but are largely underexplored because they are complex and difficult for computers to decipher. The aim of this project is to develop a digital environment that resolves this challenge and connects heterogeneous archival content to other digital sources. Within this four- year project (2016-2020), data scientists from Leiden (LIACS) and Groningen (ALICE), heritage professionals from Naturalis Biodiversity Center, historians of science from the University of Twente (STePS) and Brill publishers, work closely together. Our research is financed by the Netherlands Organization for Scientific Research (NWO) and Brill.

ERC starting grant M. Popovics / Schomaker (co-applicant), The Hands that Wrote the • Bible: Digital Palaeography and Scribal Culture of the Dead Sea Scrolls. This project aims at identifying scribes and dating documents from the Qumran scrolls using carbon-14 methods and pattern recognitions methods for characterizing writing styles in multispectral image data.

Target proeftuin: Mining Big Data, funded by SNN (co-applicant). Total budget 400k. •

External collaboration

Wiering collaborates with 1) The Center of AI Research (CAIR), Agder, Norway. A joint grant proposal has been submitted about Explainable Machine Learning, 2) UMCG, with different researchers, 3) University of Liege, Belgium, with different co-publications.

12.6 Further information

Schomaker is the co chair of the Data Science & System Complexity Center and Senior Member of the IEEE and member of the Shell/FOM programme committee on computational science & energy (CSER)

Carloni is an Associate Editor of IEEE Transactions on Robotics, of Springer-Verlag’s Journal of Intelligent Service Robotics, and of the IEEE Robotics and Automation Society

Wiering was editor of a special issue on Machine Learning for the journal “Nieuw Archief voor Wiskunde” for the September 2018 edition, together with Jim Portegies and Sander Bohte. He was in 2018 co-chair of the IEEE CIS Technical Committee on Adaptive Dynamic

189 Bernoulli Institute Annual Report

Programming and Reinforcement Learning. He was also invited in the Palace on the Dam on a Symposium on Artificial Intelligence, where he spoke with King Willem Alexander and Queen Maxima. He was also interviewed by the NRC about the topic of face recognition.

190 Bernoulli Institute Annual Report

13 Cognitive Modeling

Group leader: Prof. dr. N.A. Taatgen

Tenure track assistant professors Dr. J.P Borst RUG 1.0 Dr. F. Cnossen RUG 1.0 Dr. J.C. van Rij-Tange RUG 0.8 Dr. J.K. Spenader RUG 1.0 Dr. M.K. Van Vugt RUG 1.0

PhD students H. Berberyan RUG 1.0 (supervisor: Taatgen, Borst, van Rijn) B. Dercksen UMCG 1.0 (supervisor: Cnossen) D. Cecilio Fernandes Scholarship 1.0 (supervisor: Cnossen) M.B. Herlambang LDPD scholarship 1.0 (supervisor: Taatgen, Cnossen) C. Hoekstra PhD Scholarship 1.0 (supervisor: Taatgen, Martens) S. Huijser ERC 1.0 (supervisor: Taatgen, van Vugt) Y. Ji Self funded 1.0 (supervisor: Taatgen, van Rij) Y. Jin RUG Bursary programme 1.0 (supervisor: van Vugt, Borst) K.I. Paul Ubbo Emmius 1.0 (supervisor: Villringer, Cnossen) O. Portoles Marin DSSC scholarship 1.0 (supervisor: Cao, van Vugt) H. Shaposhnik Self funded 1.0 (supervisor: Taatgen, Borst) A.G. Toth PhD scholarship 1.0 (supervisor: Taatgen, van Rij, Hendriks)

191 Bernoulli Institute Annual Report

L.A. Newman Young Academy RUG 1.0 Bursary programme (supervisor: van Vugt) H. Yang CSC Scholarship 1.0 (supervisor: van Vugt)

Postdocs D. Bandyopadhyay RUG 1.0 (supervisor: van Vugt)

Guests Dr. M. Wirzberger, Max Planck Institute, Tubingen,¨ Germany S. Jones, Oxford, England Dr. N. Boll-Avetisyan, Potsdam, Germany Dr. S. van Ommen, Paris, France

192 Bernoulli Institute Annual Report

13.1 Research Program

The Cognitive Modeling Group studies human cognition by creating cognitive models of com- plex behavior. Cognitive models are, essentially, theories of how people think, implemented in computer simulations. To test these models, their predictions are compared to human data from behavioral and neuroimaging studies. In particular, we are interested in model-based analyses of neuroimaging data, in which a model is used to guide the interpretation of the data. Cognitive models can be applied in many domains: they can be used as the basis for design- ing education and training, or they can be used to implement intelligent agents in various applications. Cognitive Models are built using Cognitive Architectures, modeling platforms that offer support and constraints for models. Several cognitive architectures are used for modeling, including ACT-R, Nengo, and the locally developed PRIMs. A yearly international Spring School for cognitive modeling is organized in April that teaches the three central architectures, as well as error-drive learning. An overarching theme in our group is cognitive skill acquisition. We look at it in a variety of contexts, ranging from medical decision-making, to multitasking, and the transfer of skills across domains. We also investigate how people can reduce distraction and improve their mental abilities through meditation and how they acquire and interpret natural language. Last but not least, we study the use of higher-order social cognition in negotiation. In our lab we collect human data from behavioral experiments, as well as EEG and eyetracking data. We collect fMRI data and perform studies in clinical patients in collaboration with the NeuroImaging Center at the University of Groningen. We also collaborate closely with John Anderson from the Department of Psychology at Carnegie Mellon University, where we collect fMRI and MEG data. To test new user interfaces or new teaching methods, we also observe and test users, medical professionals and students in real-life settings, simulators and virtual environments, where we may also collect various psychophysiological measures such as heart rate (variability).

13.1.1 Cognitive models of persistent cognition

Multitasking People have the remarkable ability to do many things at the same time. The things that people do at the same time often interact with each other. This can lead to interference, where the performance on individual tasks suffers. But in other cases, interference is minimal or absent, and there are even cases where there is a positive effect of multitasking on performance. Our

193 Bernoulli Institute Annual Report research focuses on two questions: when does multitasking lead to interference, and why are people multitasking in the first place.

Spiking-neuron models of long- and short-term memory One crucial requirement for persistent cognition is that people can memorize information over shorter and longer intervals. To better understand how long- and short-term memory systems are implemented by our brains, we develop computational models at the level of spiking neurons. To evaluate these highly complex models with up to 1 million neurons, we analyze MEG or EEG data with machine-learning techniques, and match the results to our modelsO˜ output. In the future, these models are expected to provide ideas and input for devleoping memory systems on neuromorphic materials. This work is done in collaboration with Terrence Stewart of the University of Waterloo.

The good and the bad of mind-wandering Mind-wandering is a process that occurs abundantly and has important effects on cognitive function. On the one hand, it can hinder performance; on the other hand it can improve things such as creativity. We examine the mechanisms of mind-wandering with both behavioral data (e.g. eye-tracking, pupillometry) and cognitive models (PRIMs). We believe that combining both, fundamental research on mind wandering and modeling its mechanisms, will provide more insight in how, when, and why we mind wander. In a related line of work, we examine when mind-wandering is healthy, and when it turns into harmful rumination that is a core symptom of depression. The latter work is done in collaboration with the Neuroimaging Center (Marie-Jose van Tol), as well as Indian Institute of Technology Roorkee (India).

Cognitive training through various forms of meditation The human mind can be trained in many different ways. A way of training that currently receives a lot of attention is (mindfulness) meditation. While meditation is a popular ingredient of many courses and interventions, little is known about its cognitive mechanisms. We use ACT-R to create computational models of this form of mental training and examine how training impacts performance on cognitive tasks. In a separate line of research we also investigate the effects of reasoning-based meditation practices on cognition, emotion, and brain activity. The latter study is done in collaboration with Kent State University (United States) and Sera Jey Monastic University (India).

Multi-level architectures and Neuromorphic Computing Just like a computer architecture, a cognitive architecture can be defined and analyzed at multiple levels of abstraction. In the case of human intelligence, intelligence is implemented in the brain by neurons. To analogue in a cognitive architecture is to use neuromorphic hardware to implement properties of neurons in computer hardware in order to build energy-efficient learning networks.

194 Bernoulli Institute Annual Report

13.1.2 Model-based Neuroimaging

Understanding Problem Solving in the Brain: Mapping the Flow of Information in the Fronto-Parietal Network The key cognitive functions that are involved in problem solving have been mapped onto the fronto-parietal brain network. However, it is unclear how these crucial functions are implemented at an algorithmic level, and how the flow of information is directed within the network. In this project, we develop a detailed computational model of problem solving in the fronto-parietal network. To inform this model, we collected several EEG datasets, as well as one large fMRI dataset. Discovering cognitive processing stages using Hidden Markov Models For over a century, cognitive scientists have attempted to discover and characterize the cognitive processing stages that people go through when performing a task. For a long time, this was done using behavioral measures like reaction times. We recently developed a method that combines multivariate-pattern analysis with Hidden semi-Markov Models to automatically discover processing stages in EEG data. In this project, we are further developing this method, for example by combining it with mathematical drift diffusion models. In addition, we also apply the method to tasks of interest, for instance investigating fact learning.

Uncovering the information processing underlying the interactions between brain areas. The main aim of this project is to understand the role of oscillatory synchronization in information processing by combining cognitive modeling with a cognitive architecture and a dynamical systems analysis of EEG recorded during the same task. The EEG activity is simulated with Kuramoto models of coupled oscillators, which help to theorize about what configuration of sources could generate particular patterns of oscillatory synchronization. In collaboration with Ming Cao (control engineering).

The effect of inter-brain synchrony on social processes As two individuals collaborate, their brain activity can synchronize. In this project, we examine how inter-brain synchrony contributes to social processes in many different contacts. In one arm of the study, we examined social processes in two dancers performing together (part of the Moving Futures performing arts festival). In another arm, we examined social processes in Tibetan monks debating, as part of their reasoning-based meditation. In a more lab-based setting, we investigate how inter-brain synchronization is involved in implicit learning tasks. In collaboration with Susanne Tauber (Faculty of Economics & Business) and Ming Cao (control engineering).

195 Bernoulli Institute Annual Report

13.1.3 Skill Acquisition and Cognitive Engineering

Brain plasticity and training of psychomotor and cognitive skills in cardiovascular in- terventional medicine In this project we aim to advance our fundamental understanding of complex skill acquisition. More specifically, we want to understand how brain plasticity underlies the development of a complex skill. For this we study how medical students acquire catheter-based cardiovascular procedures. A further goal is to find out which cognitive processes can predict how well students can learn to perform such a complex medical skill. We also aim to relate this to brain changes during the acquisition of this skill. We use connective MRI data together with cognitive task performance and performance over the training sessions. This project is performed in close collaboration with the Max Planck Institute for Human Cognitive and Brain Sciences in Leipzig (Germany).

Cognitive skills in medicine: learning to perform echocardiography In this project we aim at disentangling the skill of making echocardiograms into separate knowledge domains (declarative ”what” knowledge and procedural ”how” knowledge) to understand how this skill should be taught. More specifically, we look at the interaction between these two types of knowledge, as both types have different ideal teaching methods. The main question here is how teaching medical skills should be spaced over time to promote long-term skill retention.

Serious gaming in medicine Serious games have the potential to promote better knowledge acquisition in students because they entice the student to interact with the materials longer. We apply smart learning algorithms based on a cognitive model on learning within a serious game to teach students to recognise intensive care alarms; in a related project we study how such algorithms can be used to promote learning anatomical structures. In our project on delirium recognition and treatment, on the other hand, we aim to promote better understanding in medical and nursing students of the effects of experiencing delirious episodes, which for patients are frightening and also hinder recovery resulting in longer hospital stays, or even death. To this end we assess the effects of an experiential serious game on delirium on treatment competencies and attitude changes in these students.

13.1.4 Language and Cognition

We investigate how cognitive function affects both language acquisition in childhood and language processing in adulthood. In particular we have recently focused our research on

196 Bernoulli Institute Annual Report how children and adults understand quantities, both more general quantity relationships (e.g. amounts expressed with all and each) and specific numerical quantities (e.g., two, three). Of particular interest has been how different languages code quantity, and further how executive functions (working memory, inhibition) and cognitive milestones (theory of mind development) interact with linguistic knowledge to explain quantifier interpretation.

Distributive share marking across languages In this project we investigate how different language mark plurality, focusing on distributivity in particular. Distributive share markers occur in a typologically diverse group of languages, e.g. Serbian, Korean, Japanese, Telugu and Quechua, but have hardly been investigated. This project has been the first to do controlled experimentation on distributive share markers, looking at Serbian, Korean and Telugu. In addition to developing a theoretical model of adults interpretations, the project has also looked at how the time course of quantification acquisition differs for children learning a language with distributive share markers, (e.g. Dutch) compared to children learning a language with distributive share markers (e.g. Serbian).

Distributive and collective interpretation preference across the lifespan In this project we investigate how and when people show preferences for distributive or collective interpretations, looking both at adults and children. By using experimental methods, including Dual Task experiments, we’ve been able to see the role of working memory in interpretation preferences. Our current planning will also investigate the way in which theory- of-mind abilities might influence interpretation preferences.

13.2 Research Subjects

Borst: Model-based neuroscience, neuromorphic computing Cnossen: Skill acquisition and Cognitive Engineering van Rij-Tange: Error-driven learning and Language Spenader: Language and Cognition Taatgen: Persistent Cognition, Multitasking, Neuromorphic computing van Vugt: Model-based neuroimaging, Mind Wandering Berberyan: Model-based neuroimaging Dercksen: Cognitive Engineering Cecilio Fernandes: Cognitive Engineering

197 Bernoulli Institute Annual Report

Herlambang: Cognitive Modeling, Mental Fatigue Hoekstra: Persistent Cognition Huijser: Multitasking and mind wandering Ji: Language acquistion, persistent cognition Jin: Model-based neuroimaging Paul: Cognitive engineering, neuroimaging Portoles Marin: Model-based neuroimaging Toth: Error-driven learning, Language Newman: Model-based neuroimaging Yang: Model-based neuroimaging Bandyopadhyay: Meditation, neuroimaging

13.3 Publications

13.4 Articles in scientific journals

J. Anderson, J. Borst, J. Fincham, A. Ghuman, C. Tenison, and Q. Zhang. The common • time course of memory processes revealed. Psychological Science, 29(9):1463–1474, 9 2018.

B. Arslan, R. Verbrugge, N. Taatgen, and B. Hollebrandse. Accelerating the develop- • ment of second-order false belief reasoning: A training study with different feedback methods. Child Development, 11 2018.

D. Cecilio-Fernandes, F. Cnossen, D. Jaarsma, and R. Tio. Avoiding surgical skill decay: • A systematic review on the spacing of training sessions. Journal of Surgical Education, 75(2):471–480, 4 2018. Copyright c 2017 Association of Program Directors in Surgery.

Published by Elsevier Inc. All rights reserved.

ESM-MERGE Investigators, M. van Vugt, and M. van der Velde. How does rumination • impact cognition?: A first mechanistic model. Topics in Cognitive Science, 10(1):175– 191, 1 2018.

198 Bernoulli Institute Annual Report

S. Huijser, M. van Vugt, and N. Taatgen. The wandering self: Tracking distracting self- • generated thought in a cognitively demanding context. Consciousness and Cognition, 58:170–185, 2 2018. Copyright c 2017 Elsevier Inc. All rights reserved.

W. Kennedy, M. van Vugt, and A. Banks. Editors’ introduction: Cognitive modeling • at iccm: Advancing the state of the art. Topics in Cognitive Science, 10(1):140–143, 1 2018. Copyright c 2018 Cognitive Science Society, Inc.

O. Portoles, J. Borst, and M. van Vugt. Characterizing synchrony patterns across cogni- • tive task stages of associative recognition memory. European Journal of Neuroscience, 48(8), 10 2018. This article is protected by copyright. All rights reserved. J. Spenader. Children’s comprehension of contrastive connectives. Journal of Child • Language, 45(3):610–640, 5 2018. C. Stevens, J. Daamen, E. Gaudrain, T. Renkema, J. Top, F. Cnossen, and N. Taatgen. • Using cognitive agents to train negotiation skills. Frontiers in Psychology, 9, 2 2018. N. Van Dam, M. van Vugt, D. Vago, L. Schmalzl, C. Saron, A. Olendzki, T. Meissner, • S. Lazar, J. Gorchov, K. Fox, B. Field, W. Britton, J. Brefczynski-Lewis, and D. Meyer. Reiterated concerns and further challenges for mindfulness and meditation research: A reply to davidson and dahl. Perspectives on Psychological Science, 13(1):66–69, 1 2018. N. Van Dam, M. van Vugt, D. Vago, L. Schmalzl, C. Saron, A. Olendzki, T. Meissner, • S. Lazar, C. Kerr, J. Gorchov, K. Fox, B. Field, W. Britton, J. Brefczynski-Lewis, and D. Meyer. Mind the hype: A critical evaluation and prescriptive agenda for research on mindfulness and meditation. Perspectives on Psychological Science, 13(1):36–61, 2018. R. Verbrugge, B. Meijering, S. Wierda, H. Van Rijn, and N. Taatgen. Stepwise training • supports strategic second-order theory of mind in turn-taking games. Judgement and Decision Making, 13(1):79–98, 1 2018. Q. Zhang, M. van Vugt, J. Borst, and J. Anderson. Mapping working memory retrieval • in space and in time: A combined electroencephalography and electrocorticography approach. Neuroimage, 174:472–484, 7 2018. Copyright c 2018. Published by

Elsevier Inc.

13.5 Book chapters

R. Baayen, J. van Rij, C. de Cat, and S. Wood. Autocorrelated errors in experimental • data in the language sciences: Some solutions offered by Generalized Additive Mixed

199 Bernoulli Institute Annual Report

Models, pages 49–69. Quantitative Methods in the Humanities and Social Sciences. Springer International Publishing AG, 2018. A. de Koster, J. Spenader, and P. Hendriks. Are children’s overly distributive interpreta- • tions and spreading errors related? In A. Bertolini and M. Kaplan, editors, BUCLD 42, volume 1, pages 413–426. Cascadilla Press, 5 2018. K. Paul and F. Cnossen. A Cognitive Neuroscience Perspective on Skill Acquisition in • Catheter-based Interventions. Springer Nature, 2018. J. Spenader and A. Bosnic. Distributivity preferences for dutch quantifiers elk and ieder, • 2018.

13.6 Articles in conference proceedings

A. de Koster, J. Spenader, and P. Hendriks. Child-like adults: Testing distributivity • using a dual task. 6 2018. Annual Conference on Architectures and Mechanisms for Language Processing, AMLaP ; Conference date: 06-09-2018 Through 08-11-2018. A. de Koster, J. Spenader, and P. Hendriks. The effect of working memory on distribu- • tivity interpretations and spreading errors: a dual task. 6 2018. Poster session presented at TABU Dag 2018, Groningen, Netherlands.; TABU Dag 2018 : The 39th International Linguistics Conference ; Conference date: 14-06-2018 Through 15-06-2018. A. Lieto, W. Kennedy, C. Lebiere, O. Romero, N. Taatgen, and R. West. Higher- • level knowledge, rational and social levels constraints of the common model of the mind. In A. Samsonovich and C. Lebiere, editors, Postproceedings of the 9th Annual International Conference on Biologically Inspired Cognitive Architectures, BICA 2018 (Ninth Annual Meeting of the BICA Society), held August 22-24, 2018 in Prague, Czech Republic, volume 145 of Procedia Computer Science, pages 757–764. Elsevier, 12 2018. H. Shaposhnik, N. Taatgen, and J. Borst. Predicting the optimal time for interruption • using pupillary data and classification. In Proceedings of the Annual Conference of the Cognitive Science Society, pages 2479–2484. Cognitive Science Society, 2018. S. Sprenger and J. van Rij. The development of idiom knowledge across the lifespan. 9 • 2018. Annual Conference on Architectures and Mechanisms for Language Processing, AMLaP ; Conference date: 06-09-2018 Through 08-11-2018. M. van der Velde, M. van Vugt, and N. Taatgen. Modelling the effect of depression on • working memory. In I. Juvina, J. Houpt, and C. Myers, editors, Proceedings of the 16th

200 Bernoulli Institute Annual Report

International Conference on Cognitive Modeling, page 200. University of Wisconsin, 7 2018.

R. Verbree, J. van Rij, and S. Sprenger. Cognitve effort in speech production: Insights • from pupillometry during fluent and stuttered speech. 2018. Best poster award TABU 2018.; TABU Dag 2018 : The 39th International Linguistics Conference ; Conference date: 14-06-2018 Through 15-06-2018.

R. Verbree, J. van Rij, and S. Sprenger. Mentale beanspruchung beim sprechen: • Erkenntnisse aus der pupillometrie zur flussigen¨ und gestotterten sprachproduktion. 11 2018. null ; Conference date: 24-11-2018 Through 24-11-2018.

J. Wan and N. Taatgen. The influence of music and music familiarity on time perception. • In Proceedings of the Annual Conference of the Cognitive Science Society, pages 2657–2662. Cognitive Science Society, 2018.

13.7 External funding and collaboration

Funding

Research by Jacolien van Rij is funded by an NWO-VENI grant ”Learning language from expectations”

Research by Jelmer Borst is funded by an NWO-VENI grant ”How the brain solves compli- cated problems”

Niels Taatgen and Jelmer Borst have an AFOSR funded project ”The eye is the window to working memory”

Marieke van Vugt has an AFOSR funded project “Uncovering the role of interbrain synchro- nization in social cognition.”

Niels Taatgen, together with Siewert-Jan Marrink, Shirin Faraji and Frank Noe won the CECAM-Lorentz competition to organize a Lorentz workshop on “Integrated Molecular Simulation with Machine Learning/Artificial Intelligence for Advanced Material Design”.

Marieke van Vugt was awarded an interdisciplinary PhD position “The Compassionate robot.” Young Academy of Groningen.

201 Bernoulli Institute Annual Report

Collaboration

Jelmer Borst and Niels Taatgen collaborate with John Anderson, Carnegie Mellon University

All members of the cognitive modeling group collaborate within the context of BCN (Center for Behavioral and Cognitive Neuroscience). In particular, there is a strong collaboration with Hedderik van Rijn (psychology), and Petra Hendriks (linguistics).

13.8 Further Information

The cognitive modeling group organized the yearly Spring School in Cognitive Modeling in April with approximately 30 participants

Marieke van Vugt is a member of the Young Academy Groningen

202 Bernoulli Institute Annual Report

14 Multi-Agent Systems

Group leader: Prof. dr. L.C. Verbrugge

Tenured staff (BI members) source fte Prof. dr. L.C. Verbrugge RUG 1.0 Prof. dr. H.B. Verheij RUG 1.0 Prof. dr. D. Grossi RUG 1.0

postdocs Dr. B.R.M. Gattinger (since 09-2018) FSE Fellow RUG 1.0

PhD students H. Ayoobi (since 02-2018) DSSC Marie Curie Action 1.0 (supervisors: Verheij, Verbrugge, Cao) A. Keshavarzi Zafarghandi (since 04-2018) DSSC 1.0 (supervisors: Verheij, Verbrugge, Wit) S. Pandzic Ammodo grant Kooi 1.0 (supervisors: Kooi, Verbrugge, Tamminga) Y. David Santos Ammodo grant Kooi 1.0 (supervisors: Kooi, Verbrugge) X. Su CSC 1.0 (supervisors: Kooi, Grossi, Verbrugge) J.F. van Weerden privately funded 0.5 (supervisors: Verbrugge, Hemelrijk) Y. Zhang (since 11-2018) PhD scholarship 1.0 (supervisors: Grossi, Verbrugge) H. Zheng (since 12-2018) PhD scholarship 1.0 (supervisors: Verheij, Grossi)

203 Bernoulli Institute Annual Report

Guests S. Ghosh, Indian Statistical Institute, Chennai, India Y. Wang, Peking University, Beijing, China H.P. van Ditmarsch, CNRS, Nancy, France J. Szymanik, ILLC University of Amsterdam, Amsterdam, The Netherlands S. Rey, Sorbonne University, Paris, France

204 Bernoulli Institute Annual Report

14.1 Research Program

The research group Multi-Agent Systems carries out research in the broad area of intelligent interaction, including argumentation theory, AI and law, computational social choice, multi- agent logics, and computational and logical models of social cognition. The group aims to construct new theory and computational models of higher-order social cognition and collaborative decision making in order to enhance the development of intelligent interaction between people and computer systems in ‘hybrid artificial intelligence’. The MAS group also builds computer models of argumentation in order to increase knowledge about natural argumentation, and studies the design of argumentation support software both in relation to actual task performance and in relation to formal logical models. With respect to possible applications, the research concentrates on fundamental problems from law (in particular reasoning about evidence) and the social sciences (consensus protocols for blockchain, liquid democracy, training negotiation skills). We are associate members of the Dutch graduate School for Information and Knowledge Systems (SIKS). We also participate in the school of Behavioral and Cognitive Neurosciences (BCN) and participate in the Data Science and Systems Complexity Center as well as the Groningen Center for Social Complexity Studies (GCSCS).

Logics for intelligent interaction The Multi-Agent Systems group applies logic in both well-known and novel ways. Logic provides a tested method to specify desired behavior of computational multi-agent systems. We also use epistemic logics to design communication protocols and non-monotic logics to describe legal reasoning. In our recent work, we design cognitively plausible logics for higher-order social cognition and use logical representations as evolvable forms in agent-based simulations of cognitive evolution.

Argumentation Several researchers in the group focus on research on argumentation in artificial intelligence, focusing on the connections between knowledge, data and reasoning, as a contribution to explainable, responsible and social artificial intelligence. This year, the three PhD students Ayoobi, Keshavarzi Zafarghandi and Zheng started in this direction. Ayoobi investigates argumentation in the context of home robotics (in the DSSC Cofund project). Robots must be able to handle unforeseen circumstances. New hybrid technology must be developed that combines knowledge technology for the manual representation of behavior-guiding scenarios for new and exceptional circumstances with data technology to evaluate and adapt these scenarios. In the robot architecture developed in the project, a hypothesis testing cycle will be modeled using argumentation-based techniques designed for

205 Bernoulli Institute Annual Report the combination of logic-based scenario representations and probability-based data analysis. The architecture will be tested in the international annual RoboCup@Home competition. Keshavarzi Zafarghandi (affiliated to the DSSC Cofund project) investigates argumentation and the value of data. Data is a driver of value creation. For a good understanding of the value of data, three theoretical domains are relevant: probability theory, expected utility theory and logic. Probability theory provides the foundations for descriptive statistics, expected utility theory models subjective values, and logic represents complex qualitative relations. In the project, argumentation-based formal methods and algorithms will be developed connecting probability theory, expected utility theory and logic. The project will extend techniques developed for evidential reasoning about the facts to practical reasoning about actions. Zheng (funded by the Guangzhou Elite Program) investigates argumentation for legal deci- sion support systems. The research focuses on the development of a design methodology incorporating rules, cases and arguments based on logico-probabilistic foundations. Rules describe knowledge about the patterns in the data formed by cases. Arguments are dynamically constructed and evaluated using rules that model statutes and following examples expressed in cases. Algorithms will be developed that apply rules to cases, that discover rules in cases, and that construct and evaluate arguments using rules and cases. The design methodology will be evaluated in example domains. In 2018 Davide Grossi pursued research on the definition and analysis of graduality in argumentation semantics (in cooperation with Sanjay Modgil of King’s College London, UK).

Computational social choice We pursued mainly two lines of research within computational social choice: the mathematical and computational foundations of the voting method know as Liquid Democracy, which is currently deployed by several small parties and organisations worldwide, see Figures 9 and 8. The other main line of research in this area concerns the mathematical and computational foundations of consensus protocols in blockchain (in cooperation with ILLC Amsterdam and University of Stirling, UK). In September 2018 a new PhD, Yuzhe Zhang, joined the group. Zhang’s project contributes to the line of research on Liquid Democracy.

Higher-order theory of mind This theme is concerned with higher-order social reasoning. We investigate children’s develop- ment and adults’ limitations in reasoning about the mental states of other agents, including the representations that these agents have of other agents mental states. While computer programs can correctly apply any arbitrary amount of recursion, humans frequently lose track beyond second- or third-order social reasoning. To better understand the cognitive processes involved in higher-order social cognition we are using a close-knit combination of empirical research, logic, game theory, and computational modeling.

206 Bernoulli Institute Annual Report

Figure 8: Liquid Democracy is a form of proxy voting where proxies themselves are delegable. Ultimately, the voters that decided not to delegate (gurus, or sinks in the graph) cast their ballots, weighted by the number of delegations they received, directly or indirectly. Picture source: Wikipedia.

207 Bernoulli Institute Annual Report

14.2 Overview of scientific results

Logics We showed several properties for gossip protocols, which are important in distributed comput- ing (Apt, Grossi and Van der Hoek 2018; Gattinger and Wagemakers 2018). Moreover, we showed a number of results about modeling intelligent interaction with dynamic epistemic logics and their variants (Aucher, Van Benthem and Grossi 2018; Santos 2018; Van Benthem et al. 2018) as well as default logic (Pandzic 2018). Finally, we argued that provability logic can be used to solve the paradox of the knower (De Vos et al. 2018) and we proved that provability logic enjoys a zero-one law over finite models (Verbrugge 2018).

Argumentation A first publication based on Zheng’s MSc thesis has been presented at the JURIX 2018 conference (Zheng et al., 2018). Several results on evidential reasoning were shown (Di Bello and Verheij 2018). Verheij also published the book version of his inaugural lecture ‘Arguments for Good Artificial Intelligence’ (in Dutch and in English). Verheij considers the question how we can realize the grand dreams of Artificial Intelligence, without making our worst fears come true. He argues that we need to build machines that can participate in a constructive critical discussion, that tried-and-tested tool for good science, good politics and good family life. By developing such argumentation machines we can arrive at an artificial intelligence that provides good answers to our questions, has good reasons for its actions and makes good choices. The goal is to develop argumentation systems that close the gap between knowledge-based and data-driven artificial intelligence.

Computational social choice Two matching mechanisms were shown to be strategy proof and fair (Yakiro et al. 2018; Zhang et al. 2018).

Higher-order theory of mind We showed that it is possible to accelerate the development of second-order theory of mind in five year old children by a brief training regime using stories and questions (Arslan et al. 2018), as well as to support adults in playing turn-taking games that require second-order theory of mind (Verbrugge, Meijering et al. 2018). Furthermore, we developed a statistical method to estimate the level of theory of mind that adults actually use when strategizing in a game (Diepgrond et al. 2018). Moreover, we used a strategy logic to model experimental participants’ possible reasoning strategies in turn-taking games and translated these strategy formulas into computational cognitive models to fit against human data (Ghosh and Verbrugge 2018; Top et al. 2018). Finally, we analyzed several social cognitive tasks using (parametrized) computational complexity theory (Szymanik and Verbrugge, 2018).

208 Bernoulli Institute Annual Report

0.95 1.05 tance between nodes yields a lower number of best response updates.9 This is intuitive, as larger distances between nodes 0.9 1 0.85 mean longer delegation chains, but we have not yet con- 0.95 0.8 ducted statistical tests to verify this hypothesis. 0.9 0.75 0.85 One-Shot Delegation Games 0.7 Prob. correct majority Here we study one-shot interaction in a delegation game: Mean network accuracy 0.65 0.8 all agents select their proxy (possibly themselves) simulta- 0.6 0.75 4 8 12 16 20 24 4 8 12 16 20 24 Degree Degree neously among their neighbors; no further response is pos- 0.95 1.05 sible. This contrasts the previous scenario in which agents 0.95 1.05 0.9 Random Regular Small1 World Scale Free could iteratively improve their choice based on the choices 0.9 1 0.85 0.85 0.95 of others. While Kahng, Mackenzie, and Procaccia (2018) 0.95 0.8 0.8 study a probabilistic model, we instead assume that agents 0.9 0.9 0.75 deterministically select as proxy the agent j R(i) that 0.75 0.85 2 0.7 0.85 maximizes their utility, as above. We compare q¯ and q¯⇤ (the 0.7 Prob. correct majority

Mean network accuracy 0.8 0.65 Prob. correct majority average network accuracy without and with delegation, re- Mean network accuracy 0.65 0.8 spectively), as well as the probability of a correct majority 0.6 0.75 0.6 4 8 12 16 20 24 0.75 4 8 12 16 20 24 4 8 12 16 20 24 4 8 12 16 20 24 vote under both direct democracy PD and liquid democracy Degree Degree Degree Degree PL where gurus carry as weight the number of agents for Random Regular Small World Scale Free whom they act as proxy. The difference PL PD is the no- Random Regular Small World Scale Free tion of gain (Kahng, Mackenzie, and Procaccia 2018). In Figure 9: EffectsFigure of liquid 1: Top: democracyWithout voting effort. on groupBottom: accuracy.WithTop: effortWithoutei voting effort. line with Condorcet’s jury theorem (see for instance Grof- ⇠ Bottom: With voting(0.025 effort., 0.01)Left:. Left:meanaverage accuracy network under liquid accuracy democracy. under The liq- solid (dashed) man, Owen, and Feld 1983) PD 1 as N , and in- N ! !1 line shows theuid mean democracy. (std. dev.) The of solid the initial(dashed) accuracy; line shows the dotted the mean line (std. shows maximum deed for N = 250 we obtain PD 1. First we again look at the effortless⇡ setting. Figure 1accuracy. (top) Right:dev.)probability of q; the of dotted a correct line majority shows max votei underqi. Right: delegation.probability shows both metrics for the four different network types and of a correct majority vote under liquid democracy. for different average degrees. We observe that while q¯⇤(d) 0.3 2 increases as the network degree increases (and in fact is al- 0.25 ways higher than q¯ without delegation), the probability of 1.8 0.2 a correct majority outcome, PL, simultaneously decreases. 1.6 This confirms the analysis of Kahng, Mackenzie, and Pro- 0.15 1.4 caccia (2018). We also observe that the number of gurus 0.1 decreases exponentially as the degree increases (Figure 2,

Percentage of gurus 1.2 0.05 left). Simply put, giving all voting weight to a small group Mean distance to guru 0 1 of gurus increases the chance of an incorrect majority vote, 4 8 12 16 20 24 4 8 12 16 20 24 209 assuming that gurus have a less than perfect accuracy. Degree Degree When we include effort (Figure 1, bottom), thereby mov- Random Regular Small World Scale Free ing away from the model of Kahng, Mackenzie, and Procac- Figure 2: Percentage of guru nodes under d (left) and mean cia (2018), we observe a drastic decrease in average network distance between (non-guru) nodes and their gurus (right). accuracy combined with a lower probability of a correct ma- jority outcome under liquid democracy, with both decreas- ing as network degree increases. The main reason is the ex- longer delegation chains (Figure 2, right). Intuitively, this istence of delegation cycles in this case. This contrasts the indicates that one-shot interaction in scale free networks is best response setting above where agents could iteratively more likely to end up in a local optimum. In contrast, small reconsider their choice of proxy and thus avoid cycles. Now, world networks have short average distances and thus agents even with relatively low effort (mean 0.025), up to half of all are more likely to be close to their optimal guru. agents are stuck in a cycle (and thereby fail to pass informa- tion about their type) when degree increases. This confirms results on the probability of delegation cycles from Christoff Conclusions and Future Work and Grossi (2017) and stresses the importance of cycle reso- The paper introduced delegation games as a first game- lution in concrete implementations of liquid democracy such theoretic model of liquid democracy. Both our theoretical as Liquid Feedback. and experimental results showed that voting effort is a key Finally, Figure 1 highlights differences between the four ingredient for understanding how delegations form and what network types. Scale free networks yield a lower probability their effects are. Our empirical findings provided further in- of a correct majority outcome across all degrees, as well as sights into the influence of interaction networks on the qual- a larger number of gurus with a lower average accuracy and ity of collective decisions in liquid democracy. The paper opens up several directions of research. A gen- 9More detailed results supporting this finding are presented in eral NE existence theorem is the main open question. The the supplementary material, Appendix B. framework can then be generalized along natural lines, e.g.: Bernoulli Institute Annual Report

14.3 Research subjects

H. Ayoobi: Argumentation and online learning for home robotics. B.R.M. Gattinger: Dynamic epistemic logic and communication protocols. D. Grossi: Epistemic and deontic logic, computational social choice, and argumentation. A. Keshavarzi Zafarghandi: Abstract dialectical frameworks for argumentation and dialogue games. S. Pandzic: Justification logic and default logic for belief and evidence. Y. David Santos: Dynamic epistemic and many-valued logics for belief and evidence. X. Su: Combining dynamic epistemic and deontic logic. L.C. Verbrugge: Epistemic logic and provability logic; logical and computational models of theory of mind; dynamic games. H.B. Verheij: Argumentation in artificial intelligence; AI and Law; connections between knowledge, data and reasoning. J.F. van Weerden: Individual-based models of foragers’ information exchange under preda- tion. Y. Zhang: Voting theory, especially liquid democracy. H. Zheng: Argumentation for legal decision support systems.

14.4 Publications

Books

B. Verheij. Arguments for Good Artificial Intelligence. University of Groningen, • Groningen, 2018 (book based on 2016 inaugural lecture).

Edited Books

B. Verheij and M. Wiering. Artificial Intelligence: 29th Benelux Conference, BNAIC • 2017, Groningen, The Netherlands, Revised Selected Papers. Communications in Computer and Information Science. Springer, 2018.

Contributions to books

M. Di Bello and B. Verheij. Evidential reasoning. In G. Bongiovanni, G. Postema, • A. Rotolo, G. Sartor, C. Valentini, and D. Walton, editors, Handbook of Legal Reasoning

210 Bernoulli Institute Annual Report

and Argumentation, pages 447–493. Springer, 2018. J. Szymanik and R. Verbrugge. Tractability and the computational mind. In M. Sprevak • and M. Colombo, editors, The Routledge Handbook of the Computational Mind, pages 339–354. Routledge, 2018. F. van Eemeren and B. Verheij. Argumentation theory in formal and computational • perspective. In P. Baroni, D. Gabbay, M. Giacomin, and L. van der Torre, editors, Handbook of Formal Argumentation, volume 1, pages 3–73. College Publications, 2018. B. Verheij. On coherent arguments and their inferential roles (with commentary by M. • Beirlaen). In S. Oswald and D. Maillat, editors, Argumentation and Inference, volume 1, pages 385–404. College Publications, 2018.

Articles in refereed scientific journals

B. Arslan, R. Verbrugge, N. Taatgen, and B. Hollebrandse. Accelerating the develop- • ment of second-order false belief reasoning: A training study with different feedback methods. Child Development, 2018. G. Aucher, J. van Benthem, and D. Grossi. Modal logics of sabotage revisited. Journal • of Logic and Computation, 28(2):269–303, 2018. J. van Benthem, J. van Eijck, M. Gattinger, and K. Su. Symbolic model checking for • dynamic epistemic logic–S5 and beyond. Journal of Logic and Computation, 28(2):367– 402, 2018. M. Di Bello and B. Verheij. Douglas Walton: Argument evaluation and evidence. • Argumentation, 32(2):301–307, 6 2018. S. Ghosh and R. Verbrugge. Studying strategies and types of players: Experiments, • logics and cognitive models. Synthese, 195:4265–4307, 2018. Y. David Santos, V. Bitencourt Matos, M. Moretto Ribeiro, and R. Wassermann. Partial • meet pseudo-contractions. International Journal of Approximate Reasoning, 103:11–27, 2018. J. Top, R. Verbrugge, and S. Ghosh. An automated method for building cognitive • models for turn-based games from a strategy logic. Games, 9(3), 2018. R. Verbrugge, B. Meijering, S. Wierda, H. Van Rijn, and N. Taatgen. Stepwise training • supports strategic second-order theory of mind in turn-taking games. Judgement and Decision Making, 13(1):79–98, 2018.

211 Bernoulli Institute Annual Report

H. de Weerd, D. Diepgrond, and R. Verbrugge. Estimating the use of higher-order theory • of mind using computational agents. BE Journal of Theoretical Economics, 18(2), 2018.

Editorial articles in refereed scientific journals

M. Dymitruk, R. Markovich, R. Liepin¸a, M. El Ghosh, R. van Doesburg, G. Governatori, • and B. Verheij. Research in progress: Report on the ICAIL 2017 doctoral. Artificial Intelligence and Law, 26(1):49–97, 2018.

D. Grossi and O. Roy. Introduction: Selected papers from the 4th workshop on logic, • rationality and interaction (LORI-4). Journal of Logic and Computation, 28(8):1713– 1714, 2018.

Articles in refereed conference proceedings

K. R. Apt, D. Grossi, and W. van der Hoek. When are two gossips the same? In LPAR- • 22. 22nd International Conference on Logic for Programming, Artificial Intelligence and Reasoning, Awassa, Ethiopia, 16-21 November 2018, pages 36–55, 2018.

M. Diller, A. Keshavarzi Zafarghandi, T. Linsbichler, and S. Woltran. Investigating sub- • classes of abstract dialectical frameworks. In S. Modgil, K. Budzynska, and J. Lawrence, editors, Computational Models of Argument: Proceedings of COMMA 2018, volume 305 of Frontiers in Artificial Intelligence and Applications, pages 61–72. IOS Press, 2018.

M. Gattinger and J. Wagemaker. Towards an analysis of dynamic gossip in NetKAT. • In J. Desharnais, W. Guttmann, and S. Joosten, editors, International Conference on Relational and Algebraic Methods in Computer Science, pages 280–297, Cham, 2018. Springer.

S. Pandzic. A logic of default justifications. In E. Ferme´ and S. Villata, editors, 17th • International Workshop on Nonmonotonic Reasoning (NMR 2018), pages 126–135, 2018.

Y. David Santos. A dynamic informational-epistemic logic. In A. Madeira and M. Bene- • vides, editors, Dynamic Logic, Lecture Notes in Computer Science, pages 64–81. Springer, 2018.

J. D. Top, R. Verbrugge, and S. Ghosh. Automatically translating logical strategy for- • mulas into cognitive models. In J. Houpt, I. Juvina, and C. Myers, editors, Proceedings

212 Bernoulli Institute Annual Report

of the 16th International Conference on Cognitive Modelling (ICCM 2018), pages 182–187, 2018.

R. Verbrugge. Zero-one laws with respect to models of provability logic and two • Grzegorczyk logics. In G. D’Agostino and G. Bezhanishvilii, editors, Advances in Modal Logic 2018: Accepted Short Papers, pages 115–120, 2018.

M. de Vos, B. Kooi, and R. Verbrugge. Provability logic meets the knower paradox. In • G. D’Agostino and G. Bezhanishvilii, editors, Advances in Modal Logic 2018: Accepted Short Papers, pages 31–35, 2018.

K. Yahiro, Y. Zhang, N. Barrot, and M. Yokoo. Strategyproof and fair matching • mechanism for ratio constraints. In Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, pages 59–67. International Foundation for Autonomous Agents and Multiagent Systems, 2018.

Y. Zhang, K. Yahiro, N. Barrot, and M. Yokoo. Strategyproof and fair matching • mechanism for union of symmetric m-convex constraints. In Proceedings IJCAI, pages 590–596, 2018.

H. Zheng, M. Xiong, and B. Verheij. Checking the validity of rule-based arguments • grounded in cases: A computational approach. In M. Palmirani, editor, Legal Knowledge and Information Systems. JURIX 2018: The Thirty-first Annual Conference, pages 220– 224. IOS Press, Amsterdam, 2018.

14.5 External funding and collaboration

External funding

Verbrugge acts as expert on dr. Frederik van de Putte’s Marie Curie grant ‘DYCODE’, • which was selected in 2018 and is carried out in Bayreuth.

No major grants were received by the MAS group in 2018, but Verbrugge was co- • applicant and Verheij and Grossi were members of the team that wrote the gravitation proposal “Hybrid Intelligence: Augmenting Human Intellect”, which was granted in 2019 for the period 2020-2030.

213 Bernoulli Institute Annual Report

External collaboration

Verbrugge collaborates on dynamic epistemic logics and many-valued logics with the Depart- ment of Theoretical Philosophy (prof. B.P. Kooi, dr. A. Tamminga), and with the Department of Linguistics (dr. B. Hollebrandse) and the Department of Psychology (prof. H. van Rijn) on developmental and cognitive aspects of theory of mind. She collaborates with GELIFES on agent-based models of animal cognition (prof. C.K. Hemelrijk) and with the Hanze University of Applied Sciences on agent-based models of higher-order theory of mind (dr. H.A. de Weerd). She collaborates with the ILLC, University of Amsterdam on computational models of theory of mind (dr. J. Szymanik). There is also an extensive collaboration with ISI, Chennai about logics, computational models and experiments on strategic reasoning in turn-taking games (dr. S. Ghosh).

Verheij collaborates with the City University of New York (prof. M. Di Bello) on argumentation theory and evidential reasoning. He also collaborates with the University of Amsterdam (prof. F. van Eemeren) on argumentation theory from formal and computational perspectives. Furthermore, Verheij maintains a collaboration with Sun Yat-Sen University, Guangzhou (prof. M. Xiong) on rule-based arguments grounded in cases.

Grossi collaborates with CWI Amsterdam (prof. K. Apt) and the University of Liverpool (prof. W. van der Hoek) on gossip protocols. He also coopetates with CWI Amsterdam (dr. D. Bloembergen) and TU Vienna (dr. M. Lackner) on Liquid Democracy. He maintains another line of collaboration on consensus protocols in blockchain with ILLC, University of Amsterdam (dr. R. de Haan) and the University of Stirling (Dr. A. Bracciali). Moreover, he collaborates with King’s College London (dr. S. Modgil) on argumentation semantics.

14.6 Further information

Verbrugge has a joint (‘zero’) appointment with the Department of Theoretical Philosophy. She is member of the Graduate School for Behavioral and Cognitive Neurosciences (BCN Groningen). She is board member of the Groningen Center for Social Complexity Studies. Verbrugge was a member of the BC2 of the FSE, and the search committee for several tenure trackers in the Bernoulli Institute and CogniGron. She is also a member of the Universitaire Commissie Wetenschapsbeoefening (UCW, UG) and of the Institute Advisory Board for the National Research Center for Mathematics and Computer Science (CWI) in Amsterdam. She is member of IPN (Informatics Research Platform Netherlands) and its Special Interest Group on AI. Verbrugge is elected member of the Koninklijke Hollandsche Maatschappij der Wetenschappen KHMW (Royal Holland Society of Sciences and Humanities).

214 Bernoulli Institute Annual Report

In 2018 she held several invited and keynote lectures, most notably, at the Workshop Rational Agency and Logic, Stockholm and at the Academy Colloquium on Reasoning in a Social Context, KNAW, Amsterdam. Moreover, she gave a 5-day featured tutorial at the North American Summer School on Logic, Language and Information (NASSLLI), Carnegie Mellon University, Pittsburgh. She won the prize for best AI teacher of the year and, together with Kim Veltman and Harmen de Weerd, the best poster prize of the opening symposium of the Bernoulli Institute. She is associate editor of the Journal of Logic, Language and Information and member of the editorial boards of a.o. Theoria; Computability; Journal of Philosophical Logic. She has been reviewer for a number of international journals and funding agencies and was senior programme committee member for AAMAS 2018. She was member of the search committee for a full professor position in AI at the VUA and of the promotion and tenure assessment committee for an associate professor at the Denmark Technical University in Copenhagen. She was member of the selection committee for the Nationale Wetenschaps Agenda-ORC, as well as chair of the selection committee for NWO TOP1 and TOP2 grants in mathematics, computer science and astronomy. She was member of the PhD reading committee of Maria Otworowska (Radboud University Nijmegen).

Verheij is chair of the Department of Artificial Intelligence and board member of the Bernoulli Institute. He is co-editor-in-chief of the Argument and Computation journal, as well as section editor for logic and law of the journal Artificial Intelligence and Law. Verheij is president of the IAAIL Executive Committee and vice-president of the Foundation for Legal Knowledge Systems (JURIX) and the COMMA Steering Committee as well as board member of the BNVKI. He was organising co-chair of the conference JURIX 2018 in Groningen as well as of the Facts session of the conference BNAIC 2018. Verheij gave several keynote and other invited lectures, for example, at the at the EXplainable AI in Law Workshop (XAILA 2018) at JURIX 2018, in Groningen; at the II International Congress of Law, Government and Technology in Brasilia, Brazil; and at the Stanford CodeX FutureLaw conference Legal Innovation Lightning Round. Moreover, he held the invited Orient Forum lecture at the Institute of Logic and Cognition, Zhejiang University, Hangzhou and co-lectured an invited graduate course at the Spring School on Argumentation in Arti- ficial Intelligence and Law at the Institute of Logic and Cognition, Sun Yat-Sen University, Guangzhou. As to outreach, Verheij gave an invited lecture in Dutch at the ‘Filosofisch cafe’´ in Groningen.

Grossi has a joint appointment with the Groningen Cognitive Systems and Materials Center (CogniGron). He was invited professor for a month at University Paris-Dauphine, Paris. He

215 Bernoulli Institute Annual Report was invited speaker at several conferences, for example, the Polimi Fintech Workshop in Milan; the Colloquium on Reasoning in Social Context, KNAW, Amsterdam; and the workshop From Shared Evidence to Group Attitudes, Prague. Moreover, he was invited speaker at six research seminars in the Netherlands and France. Grossi was co-organizer of the Lorentz Center Workshop on Models of Bounded Reasoning in Individuals and Groups at the Lorentz Center in Leiden. Finally, he acted as senior programme committee member of two of the main AI conferences, namely, IJCAI’18 and AAMAS’18, and programme committee member of the international conferences AAAI’18, KR’18, COMMA’18, DEON’18, LOFT’18, SR’18, ESSLLI Student Session, JURIX’18, WTSC’18.

216 Bernoulli Institute Annual Report

Colloquium Computer Science 2018 – List of Speakers

December 19 • Dr. G. Kapitsaki, University of Cyprus Licensing issues in open source software

December 11 • MsC. K. Kliffen, University of Groningen Assessing Shape Quality from Reflection Lines

December 5 • Dr. D. Taibi, University of Tampere Technical Debt and Cloud Architecture

November 28 • Dr. M. Bizzarri, University of West Bohemia Pythagorean varieties constructed from their normal spaces

October 17 • Prof. dr. O. Granmo, University of Agder The Tsetlin Machine - A Game Theoretic Bandit Driven Approach to Optimal Pattern Recognition with Propositional Logic

September 19 • Dr. J. Bernal del Nozal, University of Barcelona Intelligent Systems for Colonoscopy

July 19 • Dr. J. Ellul, University of Malta The Blockchain of Things: Challenges and Recent Work at the University of Malta

July 18 • Prof. dr. Gyan Bhanot, University of New Jersey Darwin: Everywhere and all the time

June 22 • Prof. dr. C.S. Pattichis, University of Cyprus Quantitative Imaging for Precision Medicine

June 21 • Prof. dr. C.N. Schizas, University of Cyprus EHealth for Precision Medicine

217 Bernoulli Institute Annual Report

June 19 • Dr. F. Sadikin, University of Eindhoven Intrusion Detection for IoT Systems: Hybrid method with Human-craft Rule-based Detection and Machine Learning-based Detection June 19 • Dr. J. Lee, University of Innopolis A Hybrid Trust Management Framework for Vehicular Social Networks

May 18 • Dr. S. Jansen, Basque Center of Applied Mathematics Geometric modelling and manufacturing April 12 • Dr. M. Barton, University of Utrecht Multiple Criteria Decision Support in Software Architecture Pattern Selection March 22 • Dr. Ing. S. Saralajew, University of Porche Data-driven function development: A key essential for autonomous driving

March 14 • Dr. M. Streit, University of Linz Data-driven function development: A key essential for autonomous driving January 17 • Dr. C. Neocleous - Prof. C. Schizas, University of Cyprus A fuzzy cognitive map model of macroeconomic and social parameters that influence a currency rating

218 Bernoulli Institute Annual Report

Colloquium Mathematics 2018 – List of Speakers

December 19 • Dr. M. Kool, University of Utrecht Discovering new geometry from physics

December 18 • Dr. M. Vermeeren, University of Berlin Modified equations for variational integrators

December 4 • Dr. H. Don, University of Nijmegen Synchronization of deterministic and partial automata

November 27 • Dr. F. Lucka, University College London Challenges of Mathematical Image Reconstruction

November 20 • Dr. T. Huynh, University of Brussels Strengthening Convex Relaxations of 0/1-Sets Using Boolean Formulas

Novermber 6 • Dr. N. Combe, Max Planck Institute for Mathematics Bonn Geometric invariants of the configuration space of d marked points on the complex plane

Oktober 23 • Dr. J.S. Muller,¨ University of Groningen Solving hyperelliptic diophantine equations

Oktober 16 • Dr. T. van Leeuwen, University of Utrecht Constraint-relaxation for PDE-constrained optimization in inverse problems

Oktober 9 • D. van Kekem PhD, University of Groningen Dynamics of the Lorenz-96 model

September 24 • P. Monshizadeh PhD, University of Groningen Modeling and Control of Power Systems in Microgrids

219 Bernoulli Institute Annual Report

September 20 • N. Martynshuk PhD, University of Groningen On monodromy in integrable Hamiltonian systems

September 13 • Dr. Max Lein, University of Tohoku Maxwell-type Operators — A New Class of Operators Describing Classical Waves

September 11 • Prof. T. Mountford, University of Lausanne Exponential convergence of the Frederick-Andersen model

July 2 • Yiming Bu PhD, University of Groningen A class of linear solvers based on multilevel and supernodal factorization

July 2 • Dr. M.E. Hochstenbach, University of Eindhoven Solving polynomial systems via determinantal representations” (joint work with Bor Plestenjak, Ada Boralevi, Jan Draisma, and Jasper van Doornmalen)

July 2 • Prof. dr. H.R. Bisseling, University of Utrecht Optimal 2D sparse matrix partitioning with Mondriaan” (joint work with Daan Pelt and Timon Knigge)

June 28 • E. Endo PhD, University of Groningen G-measures and external fields for long-range Ising models

June 19 • Prof. dr. J. Top, University of Groningen Rationality questions concerning Poncelet’s closure theorem

June 12 • Prof. dr. A. Dukkipati, University of Bangalore On consistency of spectral graph algorithms for community detection in networks

May 29 • Dr. V.S. Patel, University of Amsterdam Quasi Ramsey problems

220 Bernoulli Institute Annual Report

May 25 • Dr. Ir. B. Besselink, University of Groningen Model reduction for nonlinear large-scale systems May 24 • E.R. Duarte PhD, University of Groningen// Hasse-Weil Inequality and Primality Tests in the context of Curves of Genus 2 May 22 • Prof. dr. G. Lunter, University of Oxford Bayesian modeling in biological data analysis: applications to recombination and human demography May 8 • Prof. dr. G. Cornelissen, University of Utrecht Dynamics in positive characteristic April 24 • Prof. dr. A. van der Vaart, University of Leiden Bayesian uncertainty quantification April 17 • Prof. dr. D. Grieser, University of Oldenburg Geodesics on singular spaces April 10 • Prof. dr. ir. G. Jongbloed, University of Delft Steel Statistics April 6 • L. Wang PhD, University of Groningen Modeling of non-isothermal chemical reaction networks March 20 • Dr. J. Byszewski, University of Krakow Sparse generalized polynomials and nilmanifolds March 13 • Dr. R. Kang, University Nijmegen A new basic question about triangle-free graphs March 6 • Prof. dr. S. van Gils, University of Twente The Cross-Scale Effects of Neural Interactions during Human Neocortical Seizure Activity

221 Bernoulli Institute Annual Report

February 27 • Dr. C. Bertoglio, University of Groningen Boundary conditions and Lyapunov-stability in the Navier-Stokes equations February 20 • Dr. Fabian Ziltner, University of Utrecht Hamiltonian group actions on exact symplectic manifolds with proper momentum maps are standard

February 13 • Dr. Milan Bradonjic, University of New Jersey Maximum Coloring of Random Geometric Graphs January 19 • P. Behrouzi PhD, University of Wageningen Extensions of Graphical Models with Applications in Genetics and Genomics January 16// R. o´ Buachalla PhD, University of Nijmegen • - Spectral Triples and Noncommutative Fano Structures A. Krutov PhD, University of Moscou - Schubert Calculus for Quantum Grassmannians K. Strung PhD, University of Nijmegen - C*-algebras associated to quantum Grassmannians and weighted quantum Grassman- nians

222 Bernoulli Institute Annual Report

Colloquium Artificial Intelligence 2018 – List of Speakers

December 4 • Prof. dr. D. Grossi, University of Groningen Foundations for Liquid Democracy June 14 • Prof. dr. H. van Ditmarsch, University of Nancy - Asynchronous announcements Y. Wang PhD, University of Peking - Call me by your name: epistemic logic with assignments and non-rigid names June 11 • Dr. J. Szymanik, University of Amsterdam - Learnability and Semantic Universals S. Ghosh PhD, University of Chennai - Reasoning in Games

223