Objectifs de la chaire
La Chaire DeepLEVER vise à faire progresser l’état de l’art dans l’application des concepts de raisonnement automatisé, techniques et outils de raisonnement sur les modèles d’apprentissage automatique (ML).
Concrètement, nous cherchons à trouver
des explications rigoureuses quant aux prédictions faites par les modèles de ML.
Nous nous efforçons de certifier la robustesse des fonctionnements des modèles de Machine Learning.
Ce projet envisage deux axes de recherche, concrètement : l’explication et la vérification des modèles de ML profond.
Il s’appuie sur les progrès remarquables réalisés par les raisonneurs automatisés basés sur les solveurs SAT, SMT, CP, ILP (entre autres) pour une IA plus explicable et robuste basée sur les données (IA hybride pour prouver la robustesse des réseaux de neurones).
Ces réussites comptent sur une technologie de solveur améliorée, des encodages plus sophistiqués, ainsi que l’exploitation de concepts clés tels que le raffinement de l’abstraction, l’identification de la symétrie et la rupture, entre autres.
Programme : IA acceptable, certifiable & collaborative
Thèmes : développement sûr et embarquabilité, Fair learning, explicabilité, raisonnement automatique et décision
Porteur :
Joao Marquès Silva, external, University of Lisbon, PT
Co-chairs :
- Martin Cooper (UT3, IRIT)
- Emmanuel Hebrard (CNRS, LAAS)
Site
http://www.di.fc.ul.pt/~jpms/
Porteur : Joao, Marques-Silva (CNRS, IRIT)
Co-chairs :
- Martin Cooper (UT3, IRIT)
- Emmanuel Hebrard (CNRS, LAAS)
Chercheurs associés :
- Mohamed Siala (INSA Toulouse, LAAS)
- Christian Bessiere (CNRS,LIRMM)
Doctorants :
- Thomas Gerspacher, depuis avril 2020
- Xuangxiang Huang, depuis novembre 2020
Post-docs :
- Yacine Izza, depuis april 2020
Chercheurs invités :
- Gianpiero Cabodi, Politecnico di Torino, Italy, Jan 26-28 2020
-
Daniel Gibert, University of Lleida, Spain, Jan-Apr 2020
[2021]: 6 A*, 2 A
- [ICML21] Joao Marques-Silva, Thomas Gerspacher, Martin C. Cooper, Alexey Ignatiev, Nina Narodytska: Explanations for Monotonic Classifiers. ICML 2021. Preprint: https://arxiv.org/abs/2106.00154
- [IJCAI21a] Yacine Izza, Joao Marques-Silva: On Explaining Random Forests with SAT. IJCAI 2021. Preprint: https://arxiv.org/abs/2105.10278
- [IJCAI21b] Alexey Ignatiev, Joao Marques-Silva, Nina Narodytska, and Peter J. Stuckey: Reasoning- Based Learning of Interpretable ML Models. IJCAI 2021. Preprint: https://alexeyignatiev.github.io/assets/pdf/imsns-ijcai21-preprint.pdf
- [AAAI21] Alexey Ignatiev, Edward Lam, Peter J. Stuckey, Joao Marques-Silva: A Scalable Two Stage Approach to Computing Optimal Decision Sets. AAAI 2021: 3806-3814 Paper: https://ojs.aaai.org/index.php/AAAI/article/view/16498
- [KR21] Xuanxiang Huang, Yacine Izza, Alexey Ignatiev, Joao Marques-Silva: On Efficiently Explaining Graph-Based Classifiers. KR 2021. Preprint: https://arxiv.org/abs/2106.01350
- [SAT21a] Alexey Ignatiev, Joao Marques-Silva: SAT-Based Rigorous Explanations for Decision Lists. SAT 2021.Preprint: https://arxiv.org/abs/2105.06782
- [SAT21b] Stepan Kochemazov, Alexey Ignatiev, Joao Marques-Silva: Assessing Progress in SAT Solvers Through the Lens of Incremental SAT. SAT 2021. Preprint: https://alexeyignatiev.github.io/assets/pdf/kims-sat21-preprint.pdf
- [DATE21] Gianpiero Cabodi, Paolo E. Camurati, Alexey Ignatiev, Joao Marques-Silva, Marco Palena, and Paolo Pasini. DATE 2021. Preprint: https://alexeyignatiev.github.io/assets/pdf/ccimspp-date21-preprint.pdf
- [AIJ21] Martin C. Cooper, Andreas Herzig, Faustine Maffre, Frédéric Maris, Elise Perrotin, Pierre Régnier : A lightweight epistemic logic and its application to planning. Artificial Intelligence, 298 (2021).
[2020]: 6 A*, 7 A
- [NIPS20] Joao Marques-Silva, Thomas Gerspacher, Martin C. Cooper, Alexey Ignatiev, Nina Narodytska: Explaining Naive Bayes and Other Linear Classifiers with Polynomial Time and Delay. NeurIPS 2020. Paper: https://proceedings.neurips.cc/paper/2020/hash/eccd2a86bae4728b38627162ba297828-Abstract.html
- [IJCAI20a] Joao Marques-Silva, Carlos Mencía: Reasoning About Inconsistent Formulas. IJCAI 2020: 4899-4906 Paper: https://doi.org/10.24963/ijcai.2020/682
- [CP20a] Alexey Ignatiev, Martin C. Cooper, Mohamed Siala, Emmanuel Hebrard, Joao Marques-Silva: Towards Formal Fairness in Machine Learning. CP 2020: 846-867 Paper: https://doi.org/10.1007/978-3-030-58475-7_49
- [ECAI20] Oleg Zaikin, Alexey Ignatiev, João Marques-Silva: Branch Location Problems with Maximum Satisfiability. ECAI 2020: 379-386. Paper: https://doi.org/10.3233/FAIA200116
- [SAT20] Carlos Mencía, Joao Marques-Silva: Reasoning About Strong Inconsistency in ASP. SAT 2020: 332-342. Paper: https://doi.org/10.1007/978-3-030-51825-7_24
- [AIxIA20] Alexey Ignatiev, Nina Narodytska, Nicholas Asher, Joao Marques-Silva: From Contrastive to Abductive Explanations and Back Again. AI*IA 2020: 335-355 Paper: https://doi.org/10.1007/978-3-030-77091-4_21
- [ORL20] David A. Cohen, Martin C. Cooper, Artem Kaznatcheev, Mark Wallace: Steepest ascent can be exponential in bounded treewidth problems. Oper. Res. Lett. 48(3): 217-224 (2020)
- [CP20b] Martin C. Cooper: Strengthening Neighbourhood Substitution. CP 2020: 126-142
- [GKR20] David A. Cohen, Martin C. Cooper, Peter G. Jeavons, Stanislav Zivný: Galois Connections for Patterns: An Algebra of Labelled Graphs. GKR 2020: 125-150
- [IJCAI20b] Martin C. Cooper, Achref El Mouelhi, Cyril Terrioux: Variable Elimination in Binary CSPs (Extended Abstract). IJCAI 2020: 5035-5039
- [KR20] Martin C. Cooper, Andreas Herzig, Frédéric Maris, Elise Perrotin, Julien Vianey: Lightweight Parallel Multi-Agent Epistemic Planning. KR 2020: 274-283
- [STACS20] Martin C. Cooper, Simon de Givry, Thomas Schiex: Graphical Models: Queries, Complexity, Algorithms (Tutorial). STACS 2020: 4:1-4:22
- [IJCAI20c] Hao Hu, Mohamed Siala, Emmanuel Hebrard and Marie-José Huguet: Learning Optimal Decision Trees with MaxSAT and its Integration in AdaBoost. IJCAI 2020: 1170-1176 Paper: https://doi.org/10.24963/ijcai.2020/163
- [CP20c] Valentin Antuori, Emmanuel Hebrard, Marie-José Huguet, Siham Essodaigui and Alain Nguyen: Leveraging Reinforcement Learning, Constraint Programming and Local Search: A Case Study in Car Manufacturing. CP 2020: 657-672
- [AAAI20] Arthur Godet, Xavier Lorca, Emmanuel Hebrard and Gilles Simonin: Using Approximation within Constraint Programming to Solve the Parallel Machine Scheduling Problem with Additional Unit Resources. IJCAI 2020: 1170-1176 Paper: https://doi.org/10.1609/aaai.v34i02.5510
- [JAIR20] Emmanuel Hebrard and George Katsirelos. Constraint and Satisfiability Reasoning for Graph Coloring. JAIR 2020: 33-65
[2019] 3 A*, 4 A , 2 B
- [NIPS19] Alexey Ignatiev, Nina Narodytska, Joao Marques-Silva: On Relating Explanations and Adversarial Examples. NeurIPS 2019: 15857-15867 Paper: https://proceedings.neurips.cc/paper/2019/hash/7392ea4ca76ad2fb4c9c3b6a5c6e31e3-Abstract.html
- [AAAI19] Alexey Ignatiev, Nina Narodytska, João Marques-Silva: Abduction-Based Explanations for Machine Learning Models. AAAI 2019: 1511-1519 Paper: https://doi.org/10.1609/aaai.v33i01.33011511
- [IJCAI19] Alexey Ignatiev, António Morgado, Georg Weissenbacher, Joao Marques-Silva: Model- Based Diagnosis with Multiple Observations. IJCAI 2019: 1108-1115.Paper: https://doi.org/10.24963/ijcai.2019/155
- [SAT19a] Nina Narodytska, Aditya A. Shrotri, Kuldeep S. Meel, Alexey Ignatiev, Joao Marques-Silva: Assessing Heuristic Machine Learning Explanations with Model Counting. SAT 2019: 267-278 Paper: https://doi.org/10.1007/978-3-030-24258-9_19
- [SAT19b] Carlos Mencía, Oliver Kullmann, Alexey Ignatiev, Joao Marques-Silva: On Computing the Union of MUSes. SAT 2019: 211-221. Paper: https://doi.org/10.1007/978-3-030-24258-9_15 [SAT19c] António Morgado, Alexey Ignatiev, Maria Luisa Bonet, Joao Marques-Silva, Sam Buss: DRMaxSAT with MaxHS: First Contact. SAT 2019: 239-249 Paper: https://doi.org/10.1007/978-3-030- 24258-9_17
- [CPAIOR19a] Emmanuel Hebrard and George Katsirelos: A Hybrid Approach for Exact Coloring of Massive Graphs. CPAIOR 2019: 374-390
- [CPAIOR19b] Begum Genc, Mohamed Siala, Gilles Simonin and Barry O’Sullivan: An Approach to Robustness in the Stable Roommates Problem and Its Comparison with the Stable Marriage Problem.CPAIOR 2019: 320-336
- [TCS19] Begum Genc, Mohamed Siala, Gilles Simonin and Barry O’Sullivan: Complexity Study for the Robust Stable Marriage Problem. Theor. Comput. Sci. 2019: 76-92
En savoir +
Joao Marques Silva, l’IA en toute logique.
Le chercheur portugais Joao Marques Silva a posé ses valises dans la ville rose il y a deux ans. Il y poursuit ses recherches pour expliquer les décisions prises par les algorithmes. Avec l’aide de raisonnements logiques, il cherche à vérifier les modèles d’apprentissage automatique. Portrait d’un globe-trotter de l’intelligence artificielle (IA).
French 
English