4 Jul 2022
The 4th workshop on Learning & Automata - satellite of ICALP’22 https://learnaut22.github.io/ July 4th, 2022, Paris & on-line Location: building Halle aux Farines<https://u-paris.fr/batiment-de-la-halle-aux-farines/> located in the heart of the site “Grands Moulins” of the Université de Paris<https://u-paris.fr/batiment-des-grands-moulins/>. This event will be conducted in hybrid mode: in person in Paris (organizers' preferred choice) and virtually. Registration is mandatory for both modes, the registration links can be found here: https://learnaut22.github.io/registration.html It is our pleasure to inform you about LearnAut 2022, the fourth edition of the workshop, this time co-located with ICALP. Learning models defining recursive computations, like automata and formal grammars, are the core of the field called Grammatical Inference (GI). The expressive power of these models and the complexity of the associated computational problems are major research topics within mathematical logic and computer science. Historically, there has been little interaction between the GI and ICALP communities, though recently some important results started to bridge the gap between both worlds, including applications of learning to formal verification and model checking, and (co-)algebraic formulations of automata and grammar learning algorithms. The goal of this workshop is to bring together experts on logic who could benefit from grammatical inference tools, and researchers in grammatical inference who could find in logic and verification new fruitful applications for their methods. The LearnAut workshop will consist of 3 invited talks and 14 contributed talks from researchers whose submitted works were selected after a double-blind peer-reviewed phase. A significant amount of time will be kept for interactions between participants. ** Invited Speakers ** * Jeffrey Heinz, Stony Brook University, USA * Sheila McIlraith, University of Toronto, Canada * Ariadna Quattoni, Universitat Politècnica de Catalunya, Spain ** Selected papers ** * Learning state machines via efficient hashing of future traces, by Robert Baumgartner and Sicco Verwer * Towards Efficient Active Learning of PDFA, by Franz Mayr, Sergio Yovine, Federico Pan, Nicolas Basset and Thao Dang * An Algebraic Approach to Learning and Grounding, by Johanna Björklund, Adam Dahlgren Lindström and Frank Drewes * Spectral Regularization: an Inductive Bias for Sequence Modeling, by Kaiwen Hou and Guillaume Rabusseau * Marginal Inference queries in Hidden Markov Models under context-free grammar constraints, by Mohamed Reda Marzouk and Colin de la Higuera * Robust Attack Graph Generation, by Dennis Mouwen, Sicco Verwer and Azqa Nadeem * Analyzing Büchi Automata with Graph Neural Networks, by Christophe Stammet, Prisca Dotti, Ulrich Ultes-Nitsche and Andreas Fischer * Spectral Initialization of Recurrent Neural Networks: Proof of Concept, by Maude Lizaire, Simon Verret and Guillaume Rabusseau * Sequential Density Estimation via Nonlinear Continuous Weighted Finite Automata, by Tianyu Li, Bogdan Mazoure and Guillaume Rabusseau * Extending Shinohara's Algorithm for Computing Descriptive (Angluin-Style) Patterns to Subsequence Patterns, by Markus L. Schmid * Towards an AAK Theory Approach to Approximate Minimization in the Multi-Letter Case, by Clara Lacroce, Prakash Panangaden and Guillaume Rabusseau * On the limit of gradient descent for Simple Recurrent Neural Networks with finite precision, by Rémi Eyraud and Volodimir Mitarchuk * Learning regular non-deterministic distributions via non-linear optimization methods, by Wenjing Chu, Shuo Chen and Marcello Bonsangue * Learning from Positive and Negative Examples: New Proof for Binary Alphabets, by Jonas Lingg, Mateus de Oliveira Oliveira and Petra Wolf -- [LOGIC] mailing list http://www.dvmlg.de/mailingliste.html Archive: http://www.illc.uva.nl/LogicList/ provided by a collaboration of the DVMLG, the Maths Departments in Bonn and Hamburg, and the ILLC at the Universiteit van Amsterdam