About Me
Inria Rennes
LACODAM Team
Campus de Beaulieu, 263 Avenue Général Leclerc
35042 Rennes, France
E-Mail: paul.viallard@inria.fr
Links: Google Scholar, GitHub, Resume
Since October 2024, I have been working as a researcher in the LACODAM Team (led by Alexandre Termier) at Inria Rennes. From February 2024 to September 2024, I worked with Elisa Fromont and Romaric Gaudel on bandit theory as a postdoctal researcher. From February 2023 to January 2024, I worked as a postdoctal researcher in the SIERRA team (led by Francis Bach) at INRIA Paris, collaborating with Umut Şimşekli. From September 2019 to December 2022, I was a PhD student at Hubert Curien Laboratory in the Data Intelligence Team under the supervision of Emilie Morvant, Pascal Germain, and Amaury Habrard. My thesis, entitled “PAC-Bayesian Bounds and Beyond: Self-Bounding Algorithms and New Perspectives on Generalization in Machine Learning” and funded by the ANR projet APRIORI, was about providing (PAC-Bayesian) guarantees on the performance of machine learning models (such as neural networks or majority votes).
Open Positions
Internships
Publications
International Conference
Leveraging PAC-Bayes Theory and Gibbs Distributions for Generalization Bounds with Arbitrary Complexity Measure
Paul Viallard, Rémi Emonet, Amaury Habrard, Emilie Morvant, Valentina Zantedeschi
International Conference on Artificial Intelligence and Statistics (AISTATS), 2024
[pdf] [code]
Learning via Wasserstein-Based High Probability Generalisation Bounds
Paul Viallard, Maxime Haddouche, Umut Şimşekli, Benjamin Guedj
Conference on Neural Information Processing Systems (NeurIPS), 2023
[pdf] [code]
Self-bounding Majority Vote Learning Algorithms by the Direct Minimization of a Tight PAC-Bayesian C-Bound
Paul Viallard, Pascal Germain, Amaury Habrard, Emilie Morvant
European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD), 2021
[pdf] [code]
A PAC-Bayes Analysis of Adversarial Robustness
Paul Viallard, Guillaume Vidot, Amaury Habrard, Emilie Morvant
Conference on Neural Information Processing Systems (NeurIPS), 2021
[pdf] [code]
Learning Stochastic Majority Votes by Minimizing a PAC-Bayes Generalization Bound
Valentina Zantedeschi, Paul Viallard, Emilie Morvant, Rémi Emonet, Amaury Habrard, Pascal Germain, Benjamin Guedj
Conference on Neural Information Processing Systems (NeurIPS), 2021
[pdf] [code]
International Journal
A General Framework for the Practical Disintegration of PAC-Bayesian Bounds
Paul Viallard, Pascal Germain, Amaury Habrard, Emilie Morvant
Machine Learning Journal (and presented at ECML-PKDD 2023), 2023
[pdf] [code]
International Workshop
From Mutual Information to Expected Dynamics: New Generalization Bounds for Heavy-Tailed SGD
Benjamin Dupuis, Paul Viallard
NeurIPS 2023 Workshop in Heavy Tails, 2023
[pdf]
Learning via Wasserstein-Based High Probability Generalisation Bounds
Paul Viallard, Maxime Haddouche, Umut Şimşekli, Benjamin Guedj
NeurIPS 2023 Workshop Optimal Transport and Machine Learning, 2023
[pdf] [code]
Interpreting Neural Networks as Majority Votes through the PAC-Bayesian Theory
Paul Viallard, Pascal Germain, Amaury Habrard, Emilie Morvant
NeurIPS 2019 Workshop on Machine Learning with Guarantees, 2019
[pdf]
Unpublished Research Reports
Uniform Generalization Bounds on Data-Dependent Hypothesis Sets via PAC-Bayesian Theory on Random Sets
Benjamin Dupuis, Paul Viallard, George Deligiannidis, Umut Şimşekli
[pdf]
Semi-Universal Adversarial Perturbations
Jordan Frecon, Paul Viallard, Emilie Morvant, Gilles Gasso, Amaury Habrard, Stéphane Canu
[pdf]
A Theoretical Link Between Generalisation and Flat Minima
Maxime Haddouche, Paul Viallard, Umut Şimşekli, Benjamin Guedj
[pdf]
Tighter Generalisation Bounds via Interpolation
Paul Viallard, Maxime Haddouche, Umut Şimşekli, Benjamin Guedj
[pdf]
French Conferences
Bornes de généralisation: quand l’information mutuelle rencontre les bornes PAC-Bayésiennes et désintégrées
Paul Viallard
Conférence sur l’Apprentissage automatique (CAp), 2023
Intérêt des bornes désintégrées pour la généralisation avec des mesures de complexité
Paul Viallard, Rémi Emonet, Pascal Germain, Amaury Habrard, Emilie Morvant, Valentina Zantedeschi
Conférence sur l’Apprentissage automatique (CAp), 2022
Learning Stochastic Majority Votes by Minimizing a PAC-Bayes Generalization Bound
Valentina Zantedeschi, Paul Viallard, Emilie Morvant, Rémi Emonet, Amaury Habrard, Pascal Germain, Benjamin Guedj
Conférence sur l’Apprentissage automatique (CAp), 2022
Dérandomisation des Bornes PAC-Bayésiennes
Paul Viallard, Pascal Germain, Emilie Morvant
Conférence sur l’Apprentissage automatique (CAp), 2021
Apprentissage de Vote de Majorité par Minimisation d’une C-Borne PAC-Bayésienne
Paul Viallard, Pascal Germain, Emilie Morvant
Conférence sur l’Apprentissage automatique (CAp), 2021
Une Analyse PAC-Bayésienne de la Robustesse Adversariale
Guillaume Vidot, Paul Viallard, Emilie Morvant
Conférence sur l’Apprentissage automatique (CAp), 2021
Théorie PAC-Bayésienne pour l’apprentissage en deux étapes de réseaux de neurones
Paul Viallard, Rémi Emonet, Amaury Habrard, Emilie Morvant, Pascal Germain
Conférence sur l’Apprentissage automatique (CAp), 2020
Miscellaneous
PAC-Bayesian Bounds and Beyond: Self-Bounding Algorithms and New Perspectives on Generalization in Machine Learning
Paul Viallard
PhD thesis, 2022
[pdf] [slides] [code] [bibtex]
Interpreting Neural Networks as Majority Votes with the PAC-Bayesian Theory
Paul Viallard
Master thesis, 2021
[pdf] [slides]
Talks
Seminars
Aug 30th, 2024: Uniform convergence bounds via PAC-Bayes and Wasserstein distances
Journées MAS 2024, Poitiers, France
Nov 27th, 2023: Uniform Convergence in PAC-Bayesian Bounds: Wasserstein Distances and Random Sets
Data Intelligence Team Seminar, Université Jean Monnet, Saint-Etienne, France
Apr 12th, 2023: Complexity Measures in Generalization Bounds: New Results and Future Directions
OBELIX Team Seminar, Université Bretagne-Sud, online
May 9th, 2022: Towards a Practical Use of PAC-Bayesian Generalization Bounds for Learning
SIERRA Team Seminar, INRIA Paris, Paris, France
Dec 8th, 2021: A PAC-Bayes Analysis of Adversarial Robustness & Learning Stochastic Majority Votes by Minimizing a PAC-Bayes Generalization Bound
NeurIPS@Paris 2021, Sorbonne University, Paris, France
Jun 18th, 2021: Majority Vote Learning in PAC-Bayesian Theory: State of the Art and Novelty
Séminaires Traitement du Signal - Machine Learning, CNRS LIS, Aix-Marseille University, online
Science popularization
Nov 27th, 2020: La pop culture dans l’oeil des expert·es !
Nuit Européenne des Chercheur·e·s 2020, YouTube
Student Supervision
Hind Atbir — PAC-Bayesian Fair Learning
February 2024 - July 2024, co-supervised with Farah Cherfaoui, Guillaume Metzler, and Emilie Morvant
Alexiane Fraisse — Random Fourier Features and Domain Adaptation
April 2022 - July 2022, co-supervised with Guillaume Metzler, and Emilie Morvant
Luiza Dzhidzhavadze — A Multiclass C-Bound-Based Algorithm
April 2021 - June 2021, co-supervised with Emilie Morvant
Himanshu Pandey — A Multiclass C-Bound-Based Algorithm
April 2021 - June 2021, co-supervised with Emilie Morvant
Teaching
I taught during my PhD at the university of Saint-Etienne; see my resume for more details.