Curriculum Vitae - Octavian-Eugen Ganea

Postdoctoral AI researcher at MIT

CONTACT Email: oct at mit dot edu Webpage: https://people.csail.mit.edu/oct

RESEARCH INTERESTS AND VISION I am broadly interested in representation learning for unstructured data (graphs), 3D objects (e.g. molecules), text or images through statistical or geometric models that could be devised and understood in a mathematically principled and elegant manner. In particular, I explored non-Euclidean geometries in Machine Learning to overcome some of the current difficulties in graph representation learning and generation, e.g. finding and learning latent hierarchical structures in data via hyperbolic geometry, as well as combining optimal transport and graph neural networks for better models that deal with graphs. I am currently applying my models to problems related to computational chemistry such as drug discovery.

EDUCATION 2014 - 2019: PhD in Computer Science, ETH Zurich, Switzerland. • Advisor: Thomas Hofmann, Data Analytics Group • Thesis: Non-Euclidean Neural Representation Learning of Words, Entities and Hier- archies

2010 - 2012: MSc in Computer Science, Ecole Polytechnique F´ed´eralede Lausanne (EPFL), Switzerland. • MSc thesis - Google Zurich - ”Ranking entities in a geographic search engine using Bayesian networks”.

2006 - 2010: BSc in Computer Science, University Politehnica of Bucharest (UPB) • BSc thesis: Specification and Validation of a Real-Time Simple Parallel Kernel for Dependable Distributed Systems

PROFESSIONAL EXPERIENCE September 2019 - present: Postdoctoral researcher, MIT, groups of prof. Tommi Jaakkola and prof. Regina Barzilay • Graph representation learning and generation for drug discovery.

June 2018 - Oct 2018: Software engineer intern in research, Google Brain Zurich • Breaking the softmax bottleneck for text generation. Mentor: Sylvain Gelly

May 2017 - Aug 2017: Software engineer intern in research, Google Mountain View • Knowledge graph completion using deep learning and order embeddings.

2013: Research intern, Algorithms & Data Structures lab, ETH-Zurich • Algorithms for de novo peptide sequencing. Host: Peter Widmayer

Feb 2012 - Oct 2012: Software engineer intern, Google Zurich • A Bayesian network model for ranking query suggestions in Google Maps. Advisor: Radu Jurca

Jun 2011 - Sep 2011: Software engineer intern, Google Mountain View • AdSense infrastructure. RESEARCH PUBLICATIONS (in Top Tier Artificial Intelligence Conferences) 1. Computationally Tractable Riemannian Manifolds for Graph Embeddings C. Cruceru, G. B´ecigneul,O-E. Ganea Full paper at AAAI 2021: The Thirty-Fifth AAAI Conference on Artificial Intelli- gence. 2. Message Passing Networks for Molecules with Tetrahedral Chirality L. Pattanaik, O-E. Ganea, I. Coley, K. Jensen, W. Green, I. Coley Paper at NeurIPS’20 Machine Learning for Molecules Workshop. 3. Constant Curvature Graph Convolutional Networks G. Bachmann, G. B´ecigneul,O-E. Ganea Oral talk, full paper at ICML 2020: International Conference on Machine Learning. 4. Hierarchical Image Classification using Entailment Cone Embeddings A. Dhall, A. Makarova, O-E. Ganea, D. Pavllo, M. Greeff, A. Krause Intl. Workshop on Differential Geometry in Computer Vision and Machine Learning, CVPR 2020 5. Mixed-curvature Variational Autoencoders O. Skopek, O-E. Ganea, G. B´ecigneul ICLR 2020: International Conference on Learning Representations. 6. Breaking the Softmax Bottleneck via Learnable Monotonic Pointwise Non-linearities O-E. Ganea, S. Gelly, G. B´ecigneul,A. Severyn Oral talk at ICML 2019: International Conference on Machine Learning. 7. Poincar´eGloVe: Hyperbolic Word Embeddings A. Tifrea*, G. B´ecigneul*,O-E. Ganea*1 ICLR 2019: International Conference on Learning Representations. 8. Riemannian Adaptive Optimization Methods G. B´ecigneul,O-E. Ganea ICLR 2019: International Conference on Learning Representations. 9. Hyperbolic Neural Networks O-E. Ganea*, G. B´ecigneul*,T. Hofmann Spotlight (top 4% submissions) at NeurIPS 2018: Conference on Neural Infor- mation Processing Systems 10. Hyperbolic Entailment Cones for Learning Hierarchical Embeddings O-E. Ganea, G. B´ecigneul,T. Hofmann Oral talk at ICML 2018: International Conference on Machine Learning. 11. End-to-end Neural Entity Linking N. Kolitsas*, O-E. Ganea*, T. Hofmann CoNLL 2018: Conference on Natural Language Learning. 12. Learning and Evaluating Sparse Interpretable Sentence Embeddings V. Trifonov, O-E. Ganea, A. Potapenko, T. Hofmann EMNLP Workshop on the analysis and interpretation of neural networks for Natural Language Processing, 2018 13. Neural Multi-Step Reasoning for Question Answering on Semi-Structured Tables T. Haug, O-E. Ganea, P. Grnarova ECIR 2018: European Conference on Information Retrieval . 14. Web2Text: Deep Structured Boilerplate Removal T. Vogels, O-E. Ganea, C. Eickhoff ECIR 2018: European Conference on Information Retrieval . 15. Deep Joint Entity Disambiguation with Local Neural Attention O-E. Ganea, T. Hofmann EMNLP 2017: Conference on Empirical Methods in Natural Language Processing

1The * denotes equal contribution. 16. Probabilistic Bag-Of-Hyperlinks Model for Entity Linking O-E. Ganea, M. Ganea, A. Lucchi, C. Eickhoff, T. Hofmann WWW’16: International World Wide Web Conference

HONORS AND AWARDS • 2019 - Fellowship Grant - Institute for Advanced Study for the special-year program 2019 - 2020 in Machine Learning led by Sanjeev Arora. (declined) • 2010 - Excellence scholarship - Dinu Patriciu foundation (Romania) - for master studies at EPFL • 2007, 2008, 2009 - 1st and 2nd prizes - International Mathematical Contest for University Students IMC (www.imc-math.org) • 2008, 2007 - Gold Medal and Silver Medal - South Eastern Mathematical Olympiad for University Students - SEEMOUS, Greece / Cyprus • 2008, 2009 - Honorable Mention - ACM programming contest, SouthEastern European Region (http://www.acm.ro/) • 2005 - Silver Medal (3rd place) - Tuymada International Olympiad in Mathematics, Yakutsk, Russia • 2005 - Silver Medal - Balkan Mathematical Olympiad, organised in Iasi, Romania • 2001 - 2006 - one of the first 7 places every year at National Mathematical Olympiad • Top 5% in Google Code Jam programming contest, 2013.

TEACHING • October 2020: 6.867 Machine Learning lecture, MIT: guest lecture on generative models. • July 2020: guest lectures for International Mathematics Olympiad preparations for Romania’s team - https://upper.school/instructor/octavian-ganea/ • Teaching Assistant in the lectures: Deep Learning (2017, 2018 - ETHZ, Computational Intelligence Lab (2015, 2016, 2017, 2018 - ETHZ), Information Retrieval (2014, 2015, 2016 - ETHZ), Advanced Algorithms (2011 - EPFL), Graph Theory (2011 - EPFL), Concurrency (2011 - EPFL), Numerical Methods (2008-2009 - UPB) • 2007 - 2010, Lecturer for Mathematics Olympiads and Competitions and Member of the national selection committee of Romanian Mathematical Olympiad (IMO). I taught preparatory lectures to the Romanian team for the International Mathematical Olympiad. Several students of mine won prizes and medals at international contests, including gold and silver medals at IMO.

MENTORING & SUPERVISED THESES/PROJECTS: I proposed and (co-)supervised the following MSc and BSc theses (6 months projects) for the following students at ETH Zurich, some resulting in research publications (see above): 1. Panayiotou Panayiotis: Permutation Invariant Graph Generation via Optimal Trans- port, 2020 2. Octav Dragoi: Permutation Invariant Graph Generation and Optimization, 2020 3. Ondrej Skopek: Mixed-curvature Variational Autoencoders, 2019 4. Bachmann Gregor: Riemannian Graph Neural Networks, 2019 5. Andreas Bloch: Mixed-curvature Recommender Systems, 2019 6. Ankit Dhall: Hierarchical Image Captioning, 2019 7. Philipp Wirth: Hyperbolic Language Models, 2019 8. Jovan Andonov: Neural ODEs for Language Modeling, 2019 9. Calin Cruceru: Matrix Graph Embeddings, 2019 10. Alexandru Tifrea: Poincar´eGlove: Hyperbolic Word Embeddings, 2018 11. Kolitsas Nikolaos: End-to-end Neural Entity Linking, 2018 12. Valentin Trivonov: Sparse and Interpretable Sentence Embeddings 13. Igor Petrovski: Hyperbolic Sentence Embeddings, 2018 14. Junlin Yao: Detecting Medication and Adverse Drug Events from Electronic Health Records, 2018 15. Andreas Hess: Reinforcement Learning for Question Answering with Semi-structured Tables, 2017 16. Yifan Su: Deep Structured Prediction for Joint Entity Linking and Coreference Res- olution, 2016 17. Till Haug: Convolutional and Recursive Neural Networks for Question Answering on Semi-structured Tables, 2017 18. Severin Bahman: Memory Networks for Entity Linking, 2016 19. Andreas Georgiadis: Learning Sentence and Entity Representations for QA, 2016 20. Thijs Vogels: Structured Prediction for Web Page Content Extraction, 2016 21. Monteiro Jo˜aoPedro: Unsupervised Knowledge Base Fact Prediction using Matrix Word Embeddings, 2016

TALKS • October 2020: MIT, guest lecture on generative models in the 6.867 Machine Learning lecture • July 2020: Georgia Institute of Technology, host: Le Song • July 2020: Northeastern University, hosts: Tina Eliassi-Rad, Dmitri Krioukov, Rose Yu • December 2020: Bowdoin College, guest lecture, host: Jennifer Taback • April 2020: Machine Learning for Pharmaceutical Discovery and Synthesis Consor- tium, https://mlpds.mit.edu/ • March 2020: IBM Global Business Services, host: Lucia Stavarache • February 2020: University of Massachusetts Amherst, host: Andrew McCallum • January 2020:Relational.ai, host: Nikos Vasilakis • October 2019: Aggregate Intellect - https://ai.science , host: Amir Feizpour • October 2019: MIT, prof. Tommi Jaakkola’s group • October 2018: Google Brain, Zurich • April 2018: ETH Zurich Machine Learning Seminar • May 2017: Google Research Mountain View • March 2017: ETH Zurich Machine Learning Seminar • Conference talks for my papers published in NeurIPS’18, ICML’19, ICML’18, WWW’16.

COMMUNITY SERVICE • Co-founder of http://openconsulting.ai/, a free AI consulting platform for societal non-profit projects. • Reviewer for the following top tier research conferences in Artificial Intelligence: NeurIPS’20, ICML’20, AAAI’20, NeurIPS’19, EMNLP’19, ACL’18, EMNLP’18

PROGRAMMING SKILLS • Machine learning & deep learning software: PyTorch, Tensorflow • Python, Java, Lua, Scala, C++, Matlab

MISC • Finished seven mountain & road marathons • Climbed 10 four-thousand summits in the Alps - short movie of my Zinalrothorn climb • Did the longest via ferrata in Switzerland (Leukerbad) • Member of Forbes ”30 under 30”, Romania’s 2013 edition