DOCSLIB.ORG

Arxiv:2009.05908V2 [Cs.LG] 16 Jun 2021

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Arxiv:2009.05908V2 [Cs.LG] 16 Jun 2021 Understanding Boolean Function Learnability on Deep Neural Networks Anderson R. Tavares1 Pedro Avelar1 Joao˜ Flach1 Marcio Nicolau1 Lu´ıs C. Lamb1 Moshe Y. Vardi2 1Institute of Informatics, Federal University of Rio Grande do Sul, Brazil 2Department of Computer Science, Rice University, Houston, USA fartavares,phcavelar,jmflach,mnicolau,[email protected], [email protected] Abstract problems (Galassi et al. 2020; van Steenkiste et al. 2018; Kahneman et al. 2020). To respond to these challenges, Computational learning theory states that many classes of neural-symbolic methods have recently been the subject of boolean formulas are learnable in polynomial time. This intense investigation and great interest of both academic and paper addresses the understudied subject of how, in prac- tice, such formulas can be learned by deep neural networks. industry researchers (d’Avila Garcez et al. 2019; Mao et al. Specifically, we analyse boolean formulas associated with 2019; Raedt et al. 2020; Marcus 2020; Raghavan 2019). the decision version of combinatorial optimisation problems, Nobel prize winner Daniel Kahneman also referred ex- model sampling benchmarks, and random 3-CNFs with vary- plicitly to the need for symbolic reasoning systems being ing degrees of constrainedness. Our extensive experiments integrated to deep learning. At the AAAI2020 conference, indicate that: (i) regardless of the combinatorial optimisa- when Kahneman compared his own classification of the two tion problem, relatively small and shallow neural networks modes of thinking described in (Kahneman 2011), i.e.:2 Sys- are very good approximators of the associated formulas; (ii) tem 1: that operates automatically and quickly with little smaller formulas seem harder to learn, possibly due to the or no effort and no sense of voluntary control) and Sys- fewer positive (satisfying) examples available; and (iii) in- tem 2: that allocates attention to the effortful mental activ- terestingly, underconstrained 3-CNF formulas are more chal- lenging to learn than overconstrained ones. Source code and ities that demand it, including complex computations, with relevant datasets are publicly available1. the so-called AI systems 1 (deep learning) and 2 (reasoning layer), Kahneman clearly emphasized the need for a sym- bolic layer as necessary to achieve richer AI models: “...so 1 Introduction far as I’m concerned, System 1 certainly knows language... System 2... does involve certain manipulation of symbols.”. The construction of Artificial Intelligence systems that in- Moreover, at the AAAI2020 Robert S. Engelmore Memorial tegrate the fundamental cognitive abilities of reasoning and Lecture, Henry Kautz proposed several challenges for AI in learning has been pointed out by Turing Award winner and his address titled The Third AI Summer. These challenges machine learning pioneer Leslie Valiant as a key challenge include the development of effective systems along the lines for computer science (Valiant 2003, 2013). There remain, of Valiant by stating that “The next steps in AI are tighter however, several challenges with respect to closing the gap symbolic-neuro integration”3. between theoretical and practical advances in machine learn- In order to achieve such integration in AI, one has to ing that would allow for such effective integration. In the re- consider the challenges and questions still open in machine cent Montreal AI Debate between Yoshua Bengio and Gary learning and computational learning theory, such as effec- Marcus and at the recent AAAI2020 conference in New tive algorithms for reasoning and learning over classes of York, leading researchers including deep learning pioneers boolean formulas, learnable in polynomial time. Yet there arXiv:2009.05908v2 [cs.LG] 16 Jun 2021 Bengio, Hinton and LeCun have singled out the effective remains outstanding questions to be addressed, particularly development of integrated reasoning mechanisms as a key referring to effective experimentation on classes of boolean challenge to machine learning (Kahneman et al. 2020). Fur- functions (Kearns, Li, and Valiant 1994). Further, the devel- ther, machine learning in general and deep learning in par- opment of efficient learning algorithms for learning boolean ticular have achieved noticeable technological advances in a formulas remains a challenge in AI (Valiant 2013). Learn- wide range of applications. These include natural language ing unrestricted Disjunctive Normal Formulas (DNFs) still processing, machine translation, computer vision and image remains a difficult problem as proven by (Klivans and Serve- understanding, to name a few (LeCun, Bengio, and Hinton dio 2004). 2015; Schmidhuber 2015). The integration of machine learning and symbolic reason- 2Please see (Kahneman 2011) for a full account of the meaning ing has been the subject of recent debates in AI and as a of the terms and the origins of this terminology from psychology. methodology that can lead to the answer to challenging AI 3Henry Kautz AAAI2020 Robert S. Engelmore memorial lec- ture is available at https://vimeo.com/389560858. Slides are avail- 1https://github.com/machine-reasoning-ufrgs/mlbf able at https://www.cs.rochester.edu/u/kautz/talks/index.html Aiming at responding to the above challenges, this paper literals. Every BF can be transformed into a logically equiv- offers three key contributions: (i) it contributes to the inte- alent CNF; and every CNF can be converted into a logically gration of learning and reasoning, since we show that deep equivalent k-CNF (for k ≥ 3). learning can learn several families of boolean functions that Given a Boolean Formula, the Boolean Satisfiability prob- encode combinatorial optimisation problems. Regardless of lem (SAT) is to find an assignment of the variables where the the combinatorial optimisation problem, relatively small and BF evaluates to true, or to provide a proof that no satisfiying shallow neural networks are very good approximators of the assignment exists. Usually, SAT solvers takes as input a BF associated formulas; (ii) we analyse the difficulty of learning in the CNF format. The SAT problem is important because it boolean formulas of varying sizes and contrainedness over can be found in different areas in science and also in numer- the clauses in CNF formulations. Our extensive experiments ous practical problems. Besides that, it is general enough show that smaller formulas can be harder to learn, possi- that several other problems that, at first, are not related to bly due to the fewer positive (satisfying) examples available; SAT, can be converted into a SAT problem - for example, and (iii) interestingly, underconstrained 3-CNF formulas are some problems in graph theory, such as the clique and col- more challenging to learn than overconstrained ones. oring problems (Garey and Johnson 1979). The remainder of this paper is organised as follows. Sec- tion 2 presents the fundamental concepts and definitions about boolean learnability and related work in the field. Related work Computational learning theory presents Section 3 presents our methodology to assess the learn- hardness results on the learnability of boolean functions re- ing capabilities of deep neural networks on boolean func- lated to certain classes of problems, such as cryptography tions. Section 4 presents our learnability experiments on (Rivest 1991), robust learning (Gourdeau et al. 2019) and boolean formulas encoding combinatorial optimisation for- distribution learning (Kearns et al. 1994). It also presents mulas, whereas Section 5 investigates random 3-CNFs with many positive results on polynomial-time learnability of various number of variables and clause-to-variable ratios boolean formulas, which are of our interest. Here we (constrainedness). Section 6 presents concluding remarks describe results concerning conjunctive normal formulas and directions for future work. (CNFs) and neural networks. Within the probably approximately correct (PAC) learn- 2 On Deep Boolean Function Learnability ing framework, (Valiant 1984) shows that conjunctive nor- Already in his seminal PAC learning paper (Valiant 1984), mal formulas with a bounded number of literals per clause Valiant highlighted the importance of knowledge represen- (k-CNFs) are learnable in general, although not mentioning tation and the relationship with logic and the design of ma- neural networks specifically. chine learning systems: “...[the] remaining design choice Artificial Neural networks (ANNs) are universal learn- that has to be made is that of knowledge representation. ers of boolean formulas (Blum 1989; Steinbach and Kohut Since our declared aim is to represent general knowledge, 2002), since classical perceptrons can be arranged to im- it seems almost unavoidable that we use some kind of logic plement any logical gate and such gates can be arranged rather than, for example, formal grammars or geometri- to implement any boolean formula, also with the possi- cal constructs. [...] we shall represent concepts as Boolean blity of extracting boolean formulas from trained neural functions of a set of propositional variables. The recogni- networks (Tsukimoto 1997). Moreover, even single-hidden- tion algorithms that we attempt to deduce will be therefore layer networks are universal boolean function learners (An- Boolean circuits or expressions.” Over the years, machine thony 2010), although the worst-case number of neurons in learning and automated reasoning (in A.I., knowledge rep- the hidden layer is exponential on the number of inputs. (An-
Load more

Recommended publications
  • Computational Learning Theory: New Models and Algorithms
    Computational Learning Theory: New Models and Algorithms by Robert Hal Sloan S.M. EECS, Massachusetts Institute of Technology (1986) B.S. Mathematics, Yale University (1983) Submitted to the Department- of Electrical Engineering and Computer Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 1989 @ Robert Hal Sloan, 1989. All rights reserved The author hereby grants to MIT permission to reproduce and to distribute copies of this thesis document in whole or in part. Signature of Author Department of Electrical Engineering and Computer Science May 23, 1989 Certified by Ronald L. Rivest Professor of Computer Science Thesis Supervisor Accepted by Arthur C. Smith Chairman, Departmental Committee on Graduate Students Abstract In the past several years, there has been a surge of interest in computational learning theory-the formal (as opposed to empirical) study of learning algorithms. One major cause for this interest was the model of probably approximately correct learning, or pac learning, introduced by Valiant in 1984. This thesis begins by presenting a new learning algorithm for a particular problem within that model: learning submodules of the free Z-module Zk. We prove that this algorithm achieves probable approximate correctness, and indeed, that it is within a log log factor of optimal in a related, but more stringent model of learning, on-line mistake bounded learning. We then proceed to examine the influence of noisy data on pac learning algorithms in general. Previously it has been shown that it is possible to tolerate large amounts of random classification noise, but only a very small amount of a very malicious sort of noise.
    [Show full text]
  • On Computational Tractability for Rational Verification
    Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI-19) On Computational Tractability for Rational Verification Julian Gutierrez1 , Muhammad Najib1 , Giuseppe Perelli2 , Michael Wooldridge1 1Department of Computer Science, University of Oxford, UK 2Department of Informatics, University of Leicester, UK fjulian.gutierrez, mnajib, [email protected], [email protected] Abstract theoretic (e.g., Nash) equilibrium. Unlike model checking, rational verification is still in its infancy: the main ideas, Rational verification involves checking which formal models, and reasoning techniques underlying rational temporal logic properties hold of a concur- verification are under development, while current tool sup- rent/multiagent system, under the assumption that port is limited and cannot yet handle systems of industrial agents in the system choose strategies in game the- size [Toumi et al., 2015; Gutierrez et al., 2018a]. oretic equilibrium. Rational verification can be un- derstood as a counterpart of model checking for One key difficulty is that rational verification is computa- multiagent systems, but while model checking can tionally much harder than model checking, because checking be done in polynomial time for some temporal logic equilibrium properties requires quantifying over the strategies specification languages such as CTL, and polyno- available to players in the system. Rational verification is also mial space with LTL specifications, rational ver- different from model checking in the kinds of properties that ification is much more intractable: 2EXPTIME- each technique tries to check: while model checking is inter- any complete with LTL specifications, even when using ested in correctness with respect to possible behaviour of explicit-state system representations.
    [Show full text]
  • Downloaded from 128.205.114.91 on Sun, 19 May 2013 20:14:53 PM All Use Subject to JSTOR Terms and Conditions 660 REVIEWS
    Association for Symbolic Logic http://www.jstor.org/stable/2274542 . Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Association for Symbolic Logic is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Symbolic Logic. http://www.jstor.org This content downloaded from 128.205.114.91 on Sun, 19 May 2013 20:14:53 PM All use subject to JSTOR Terms and Conditions 660 REVIEWS The penultimate chapter, Real machines, is the major exposition on Al techniques and programs found in this book. It is here that heuristic search is discussed and classic programs such as SHRDLU and GPS are described. It is here that a sampling of Al material and its flavor as research is presented. Some of the material here is repeated without real analysis. For example, the author repeats the standard textbook mistake on the size of the chess space. On page 178, he states that 10120 is the size of this space, and uses this to suggest that no computer will ever play perfect chess. Actually, an estimate of 1040 is more realistic. If one considers that no chess board can have more than sixteen pieces of each color and there are many configurations that are illegal or equivalent, then the state space is reduced considerably.
    [Show full text]
  • Logics for Concurrency
    Lecture Notes in Computer Science 1043 Edited by G. Goos, J. Hartmanis and J. van Leeuwen Advisory Board: W. Brauer D. Gries J. Stoer Faron Moiler Graham Birtwistle (Eds.) Logics for Concurrency Structure versus Automata ~ Springer Series Editors Gerhard Goos, Karlsruhe University, Germany Juris Hartmanis, Cornetl University, NY, USA Jan van Leeuwen, Utrecht University, The Netherlands Volume Editors Faron Moller Department ofTeleinformatics, Kungl Tekniska H6gskolan Electrum 204, S-164 40 Kista, Sweden Graham Birtwistle School of Computer Studies, University of Leeds Woodhouse Road, Leeds LS2 9JT, United Kingdom Cataloging-in-Publication data applied for Die Deutsche Bibliothek - CIP-Einheitsaufnahme Logics for concurrency : structure versus automata / Faron Moller; Graham Birtwistle (ed.). - Berlin ; Heidelberg ; New York ; Barcelona ; Budapest ; Hong Kong ; London ; Milan ; Paris ; Santa Clara ; Singapore ; Tokyo Springer, 1996 (Lecture notes in computer science ; Voi. 1043) ISBN 3-540-60915-6 NE: Moiler, Faron [Hrsg.]; GT CR Subject Classification (1991): F.3, F.4, El, F.2 ISBN 3-540-60915-6 Springer-Verlag Berlin Heidelberg New York This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer -Verlag. Violations are liable for prosecution under the German Copyright Law. Springer-Verlag Berlin Heidelberg 1996 Printed in Germany Typesetting: Camera-ready by author SPIN 10512588 06/3142 - 5 4 3 2 1 0 Printed on acid-free paper Preface This volume is a result of the VIII TH BANFF HIGHER ORDER WORKSHOP held from August 27th to September 3rd, 1994, at the Banff Centre in Banff, Canada.
    [Show full text]
  • Conversational Concurrency Copyright © 2017 Tony Garnock-Jones
    CONVERSATIONALCONCURRENCY tony garnock-jones Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy College of Computer and Information Science Northeastern University 2017 Tony Garnock-Jones Conversational Concurrency Copyright © 2017 Tony Garnock-Jones This document was typeset on December 31, 2017 at 9:22 using the typographical look-and-feel classicthesis developed by André Miede, available at https://bitbucket.org/amiede/classicthesis/ Abstract Concurrent computations resemble conversations. In a conversation, participants direct ut- terances at others and, as the conversation evolves, exploit the known common context to advance the conversation. Similarly, collaborating software components share knowledge with each other in order to make progress as a group towards a common goal. This dissertation studies concurrency from the perspective of cooperative knowledge-sharing, taking the conversational exchange of knowledge as a central concern in the design of concur- rent programming languages. In doing so, it makes five contributions: 1. It develops the idea of a common dataspace as a medium for knowledge exchange among concurrent components, enabling a new approach to concurrent programming. While dataspaces loosely resemble both “fact spaces” from the world of Linda-style lan- guages and Erlang’s collaborative model, they significantly differ in many details. 2. It offers the first crisp formulation of cooperative, conversational knowledge-exchange as a mathematical model. 3. It describes two faithful implementations of the model for two quite different languages. 4. It proposes a completely novel suite of linguistic constructs for organizing the internal structure of individual actors in a conversational setting. The combination of dataspaces with these constructs is dubbed Syndicate.
    [Show full text]
  • Symposium on Principles of Database Systems Paris, France – June 14-16, 2004
    23ndACM SIGMOD-SIGACT- SIGART Symposium on Principles of Database Systems Paris, France – June 14-16, 2004 http://www.sciences.univ-nantes.fr/irin/SIGMODPODS04/ The symposium invites papers on fundamental aspects of data management. Original research papers on the theory, design, specification, or implementation of data management tools are solicited. Papers emphasizing new topics or foundations of emerging areas are especially welcome. The symposium will be held at the Maison de la Chimie, Paris, in conjunction with the ACM SIGMOD Intl. Conference on Management of Data (SIGMOD’04). Suggested topics include the following (this list is not exhaustive and the order does not reflect priorities): Access Methods & Physical Design Databases & Workflows Real-time Databases Active Databases Deductive Databases & Knowledge Bases Security & Privacy Complexity & Performance Evaluation Distributed Databases Semistructured Data & XML Data Integration & Interoperability Information Processing on the Web Spatial & Temporal Databases Data Mining Logic in Databases Theory of recovery Data Models Multimedia Databases Transaction Management Data Stream Management Object-oriented Databases Views & Warehousing Database Programming Languages Query Languages Web Services & Electronic Commerce Databases & Information Retrieval Query Optimization XML Databases Important Dates: Paper abstracts due: December 1, 2003 (11:59pm PST) Full papers due: December 8, 2003 (11:59pm PST) Notification of acceptance/rejection: February 23, 2004 Camera-ready due: March 22, 2004
    [Show full text]
  • Self-Sorting SSD: Producing Sorted Data Inside Active Ssds
    Foreword This volume contains the papers presentedat the Third Israel Symposium on the Theory of Computing and Systems (ISTCS), held in Tel Aviv, Israel, on January 4-6, 1995. Fifty five papers were submitted in response to the Call for Papers, and twenty seven of them were selected for presentation. The selection was based on originality, quality and relevance to the field. The program committee is pleased with the overall quality of the acceptedpapers and, furthermore, feels that many of the papers not used were also of fine quality. The papers included in this proceedings are preliminary reports on recent research and it is expected that most of these papers will appear in a more complete and polished form in scientific journals. The proceedings also contains one invited paper by Pave1Pevzner and Michael Waterman. The program committee thanks our invited speakers,Robert J. Aumann, Wolfgang Paul, Abe Peled, and Avi Wigderson, for contributing their time and knowledge. We also wish to thank all who submitted papers for consideration, as well as the many colleagues who contributed to the evaluation of the submitted papers. The latter include: Noga Alon Alon Itai Eric Schenk Hagit Attiya Roni Kay Baruch Schieber Yossi Azar Evsey Kosman Assaf Schuster Ayal Bar-David Ami Litman Nir Shavit Reuven Bar-Yehuda Johan Makowsky Richard Statman Shai Ben-David Yishay Mansour Ray Strong Allan Borodin Alain Mayer Eli Upfal Dorit Dor Mike Molloy Moshe Vardi Alon Efrat Yoram Moses Orli Waarts Jack Feldman Dalit Naor Ed Wimmers Nissim Francez Seffl Naor Shmuel Zaks Nita Goyal Noam Nisan Uri Zwick Vassos Hadzilacos Yuri Rabinovich Johan Hastad Giinter Rote The work of the program committee has been facilitated thanks to software developed by Rob Schapire and FOCS/STOC program chairs.
    [Show full text]
  • COMP 516 Research Methods in Computer Science Research Methods in Computer Science Lecture 3: Who Is Who in Computer Science Research
    COMP 516 COMP 516 Research Methods in Computer Science Research Methods in Computer Science Lecture 3: Who is Who in Computer Science Research Dominik Wojtczak Dominik Wojtczak Department of Computer Science University of Liverpool Department of Computer Science University of Liverpool 1 / 24 2 / 24 Prizes and Awards Alan M. Turing (1912-1954) Scientific achievement is often recognised by prizes and awards Conferences often give a best paper award, sometimes also a best “The father of modern computer science” student paper award In 1936 introduced Turing machines, as a thought Example: experiment about limits of mechanical computation ICALP Best Paper Prize (Track A, B, C) (Church-Turing thesis) For the best paper, as judged by the program committee Gives rise to the concept of Turing completeness and Professional organisations also give awards based on varying Turing reducibility criteria In 1939/40, Turing designed an electromechanical machine which Example: helped to break the german Enigma code British Computer Society Roger Needham Award His main contribution was an cryptanalytic machine which used Made annually for a distinguished research contribution in computer logic-based techniques science by a UK based researcher within ten years of their PhD In the 1950 paper ‘Computing machinery and intelligence’ Turing Arguably, the most prestigious award in Computer Science is the introduced an experiment, now called the Turing test A. M. Turing Award 2012 is the Alan Turing year! 3 / 24 4 / 24 Turing Award Turing Award Winners What contribution have the following people made? The A. M. Turing Award is given annually by the Association for Who among them has received the Turing Award? Computing Machinery to an individual selected for contributions of a technical nature made Frances E.
    [Show full text]
  • Adventures in Monotone Complexity and TFNP
    Adventures in Monotone Complexity and TFNP Mika Göös1 Institute for Advanced Study, Princeton, NJ, USA [email protected] Pritish Kamath2 Massachusetts Institute of Technology, Cambridge, MA, USA [email protected] Robert Robere3 Simons Institute, Berkeley, CA, USA [email protected] Dmitry Sokolov4 KTH Royal Institute of Technology, Stockholm, Sweden [email protected] Abstract Separations: We introduce a monotone variant of Xor-Sat and show it has exponential mono- tone circuit complexity. Since Xor-Sat is in NC2, this improves qualitatively on the monotone vs. non-monotone separation of Tardos (1988). We also show that monotone span programs over R can be exponentially more powerful than over finite fields. These results can be interpreted as separating subclasses of TFNP in communication complexity. Characterizations: We show that the communication (resp. query) analogue of PPA (subclass of TFNP) captures span programs over F2 (resp. Nullstellensatz degree over F2). Previously, it was known that communication FP captures formulas (Karchmer–Wigderson, 1988) and that communication PLS captures circuits (Razborov, 1995). 2012 ACM Subject Classification Theory of computation → Communication complexity, The- ory of computation → Circuit complexity, Theory of computation → Proof complexity Keywords and phrases TFNP, Monotone Complexity, Communication Complexity, Proof Com- plexity Digital Object Identifier 10.4230/LIPIcs.ITCS.2019.38 Related Version A full version of the paper is available at [24], https://eccc.weizmann.ac. il/report/2018/163/. Acknowledgements We thank Ankit Garg (who declined a co-authorship) for extensive discus- sions about monotone circuits. We also thank Thomas Watson and anonymous ITCS reviewers for comments. 1 Work done while at Harvard University; supported by the Michael O.
    [Show full text]
  • CRN Equal Access to Financial Aid and a Continued on Page 8 Best Practices Memo
    COMPUTING RESEARCH NEWS A Publication of the Computing Research Association September 2003 Vol. 15/No. 4 Guest Column Title IX and Women in Academics By Senator Ron Wyden in traditionally male-dominated the right thing to do; it is the smart fields such as the hard sciences, math thing to do. This fall the athletic fields of and engineering—disciplines where The numbers reveal a striking America’s elementary and secondary our nation needs competent workers inequity when it comes to gender schools, colleges and universities will now more than ever before. representation in the math, science resound with the voices of girls and We can all agree that fairness and technology fields. A National young women who choose to include implores us to create and enforce Science Foundation study found that sports as part and parcel of their equal opportunity for women in women accounted for only 23 per- educational experience. Those girls math, science and technology. That cent of physical scientists and 10 and young women will not only be is a compelling argument in itself, percent of engineers. The percent- taking physical exercise; they’ll be but it is not the only argument. A ages of women on faculties in these exercising their rights to equal report from the Hart-Rudman areas are even lower, with 14 percent Senator Ron Wyden opportunity under a law known as Commission on National Security to of science faculty members being Title IX. 2025 warned that America’s failure women and a mere 6 percent in Title IX states a simple principle.
    [Show full text]
  • CS256/Spring 2008 — Lecture #11 Zohar Manna
    CS256/Spring 2008 — Lecture #11 Zohar Manna 11-1 Beyond Temporal Logics Temporal logic expresses properties of infinite sequences of states, but there are interesting properties that cannot be expressed, e.g., “p is true only (at most) at even positions.” Questions (foundational/practical): • What other languages can we use to express properties of sequences (⇒ properties of programs)? • How do their expressive powers compare? • How do their computational complexities (for the decision problems) compare? 11-2 ω-languages Σ: nonempty finite set (alphabet) of characters Σ∗: set of all finite strings of characters in Σ finite word w ∈ Σ∗ Σω: set of all infinite strings of characters in Σ ω-word w ∈ Σω (finitary) language: L ⊆ Σ∗ ω-language: L ⊆ Σω 11-3 States Propositional LTL (PLTL) formulas are constructed from the following: • propositions p1,p2,...,pn. • boolean/temporal operators. • a state s ∈{f, t}n i.e., every state s is a truth-value assignment to all n propositional variables. Example: If n = 3, then s : hp1 : t, p2 : f, p3 : ti corresponds to state tft. p1 ↔ p2 denotes the set of states {fff,fft,ttf,ttt} • alphabet Σ = {f, t}n i.e, 2n strings, one string for every state. Note: t, f = formulas (syntax) 11-4 t, f = truth values (semantics) Models of PLTL 7→ ω-languages • A model of PLTL for the language with n propositions σ : s0,s1,s2,... can be viewed as an infinite string s0s1s2 . , i.e., σ ∈ ({f, t}n)ω • A PLTL formula ϕ denotes an ω-language n ω L = {σ | σ Õ ϕ} ⊆ ({f, t} ) Example: If n = 3, then ϕ : ¼ (p1 ↔ p2) denotes the ω-language L(ϕ) = {fff,fft,ttf,ttt}ω 11-5 Other Languages to Talk about Infinite Sequences • ω-regular expressions • ω-automata 11-6 Regular Expressions Syntax: ∗ r ::= ∅ | ε | a | r1r2 | r1 + r2 | r (ε = empty word, a ∈ Σ) Semantics: A regular expression r (on alphabet Σ) denotes a finitary language L(r) ⊆ Σ∗: L(∅) = ∅ L(ε) = {ε} L(a) = {a} L(r1r2) = L(r1) · L(r2) = {xy | x ∈ L(r1), y ∈ L(r2)} L(r1 + r2) = L(r1) ∪ L(r2) ∗ ∗ L(r ) = L(r) = {x1x2 · · · xn | n ≥ 0, x1, x2, .
    [Show full text]
  • 30Th International Symposium on Theoretical Aspects of Computer Science
    30th International Symposium on Theoretical Aspects of Computer Science STACS’13, February 27th to March 2nd, 2013, Kiel, Germany Edited by Natacha Portier Thomas Wilke LIPIcs – Vol. 20 – STACS’13 www.dagstuhl.de/lipics Editors Natacha Portier Thomas Wilke École Normale Supérieure de Lyon Christian-Albrechts-Universität zu Kiel Lyon Kiel [email protected] [email protected] ACM Classification 1998 F.1.1 Models of Computation, F.2.2 Nonnumerical Algorithms and Problems, F.4.1 Mathematical Logic, F.4.3 Formal Languages, G.2.1 Combinatorics, G.2.2 Graph Theory ISBN 978-3-939897-50-7 Published online and open access by Schloss Dagstuhl – Leibniz-Zentrum für Informatik GmbH, Dagstuhl Publishing, Saarbrücken/Wadern, Germany. Online available at http://www.dagstuhl.de/dagpub/978-3-939897-50-7. Publication date February, 2013 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. License This work is licensed under a Creative Commons Attribution-NoDerivs 3.0 Unported license (CC-BY- ND 3.0): http://creativecommons.org/licenses/by-nd/3.0/legalcode. In brief, this license authorizes each and everybody to share (to copy, distribute and transmit) the work under the following conditions, without impairing or restricting the authors’ moral rights: Attribution: The work must be attributed to its authors. No derivation: It is not allowed to alter or transform this work. The copyright is retained by the corresponding authors. Digital Object Identifier: 10.4230/LIPIcs.STACS.2013.i ISBN 978-3-939897-50-7 ISSN 1868-8969 http://www.dagstuhl.de/lipics iii LIPIcs – Leibniz International Proceedings in Informatics LIPIcs is a series of high-quality conference proceedings across all fields in informatics.
    [Show full text]