Professeur adjoint en informatique
Assistant professor in computer science

Département d’informatique et de génie logiciel, Pavillon Adrien-Pouliot, Local PLT-3949
Université Laval, Québec (QC), Canada, G1V 0A6

Courriel / e-mail: firstname.lastname@ift.ulaval.ca

Chercheur en apprentissage automatique
Machine learning researcher

Intérêts de recherche / Research interests:
Statistical machine learning theory (with an emphasis on PAC-Bayesian learning), domain adaptation, learning algorithms, representation learning, interpretability ...

Students and Postdocs

Current
  • Benjamin Leblanc (PhD candidate)
  • Sokhna Diarra Mbacke (PhD candidate)
  • Thibaud Godon (PhD candidate, co-supervised by Alexandre Drouin and Jacques Corbeil)
  • Louis-Philippe Vignault (MSc candidate, co-supervised with Audrey Durand)
  • Élina Francovic-Fontaine (PhD candidate, co-supervised with Jacques Corbeil and Elsa Rousseau)
  • Sandrine Blais-Deschênes (PhD candidate co-supervised with Josée Desharnais)
  • Mathieu Bazinet (PhD candidate, co-supervised by Valentina Zantedeschi)
  • Mariame Gnéré Coulibaly (MSc candidate, co-supervised with Elsa Rousseau)
  • Simon Bertrand (PhD candidate, co-supervised with Nadia Tawbi)
  • Shubham Gupta (PhD candidate co-supervised with Cem Subakan)
  • Étienne Gael Tajeuna (postdoc, co-supervised by Jacques Corbeil)
Past
  • Gaël Letarte (PhD co-supervised by François Laviolette)
  • Paul Viallard (PhD co-supervised with Amaury Habrard and Emilie Morvant)
  • Mathieu Alain (Master thesis co-supervised with François Laviolette)
  • Luxin Zhang (PhD co-supervised with Christophe Biernacki and Yacine Kessaci)
  • Vera Shalaeva (postdoc researcher)




Twitter, Linkedin, GitHub





Publications

See also: Google Scholar, dblp, Semantic Scholar, arXiv
Selected Reports
A Greedy Algorithm for Building Compact Binary Activated Neural Networks [ arXiv ]
Benjamin Leblanc, Pascal Germain (2022)

Implicit Variational Inference: the Parameter and the Predictor Space [ arXiv ] [ bibtex ]
Yann Pequignot, Mathieu Alain, Patrick Dallaire, Alireza Yeganehparast, Pascal Germain, Josée Desharnais, François Laviolette (2020)

Peer-Reviewed Works (conference, journals, workshops)
Statistical Guarantees for Variational Autoencoders using PAC-Bayesian Theory
Sokhna Diarra Mbacke, Florence Clerc, Pascal Germain (NeurIPS 2023)

PAC-Bayesian Generalization Bounds for Adversarial Generative Models [ proceedings ] [ arXiv ] [ bibtex ]
Sokhna Diarra Mbacke, Florence Clerc, Pascal Germain (ICML 2023)

Invariant Causal Set Covering Machines [ paper ] [ bibtex ]
Thibaud Godon, Baptiste Bauvin, Pascal Germain, Jacques Corbeil, Alexandre Drouin (ICML 2023 Workshop on Spurious Correlations, Invariance, and Stability)

Sample Boosting Algorithm (SamBA) - An Interpretable Greedy Ensemble Classifier Based On Local Expertise For Fat Data [ proceedings ] [ bibtex ]
Baptiste Bauvin, Cécile Capponi, Florence Clerc, Pascal Germain, Sokol Koço, Jacques Corbeil (UAI 2023)

PAC-Bayesian Learning of Aggregated Binary Activated Neural Networks with Probabilities over Representations [ proceedings ] [ arXiv ] [ bibtex ]
Louis Fortier-Dubois, Benjamin Leblanc, Gaël Letarte, François Laviolette, Pascal Germain (CANAI 2023)

A General Framework for the Practical Disintegration of PAC-Bayesian Bounds [ arXiv preprint ]
Paul Viallard, Pascal Germain, Amaury Habrard, Emilie Morvant (Mach. Learn. 2023)

Interpretable Domain Adaptation Using Unsupervised Feature Selection on Pre-trained Source Models [ paper ] [ preprint ] [ code ]
Luxin Zhang, Pascal Germain, Yacine Kessaci, Christophe Biernacki (Neurocomputing 2022)

Interpretable Domain Adaptation for Hidden Subdomain Alignment in the Context of Pre-Trained Source Models [ proceedings ] [ supplemental ] [ HAL ] [ spotlight ] [ code ]
Luxin Zhang, Pascal Germain, Yacine Kessaci, Christophe Biernacki (AAAI 2022)

Learning Stochastic Majority Votes by Minimizing a PAC-Bayes Generalization Bound [ proceedings ] [ arXiv ]
Valentina Zantedeschi, Paul Viallard, Emilie Morvant, Rémi Emonet, Amaury Habrard, Pascal Germain, Benjamin Guedj (NeurIPS 2021)

Self-Bounding Majority Vote Learning Algorithms by the Direct Minimization of a Tight PAC-Bayesian C-Bound [ paper ]
Paul Viallard, Pascal Germain, Amaury Habrard, Emilie Morvant (ECML 2021)

Target to Source Coordinate-wise Adaptation of Pre-trained Models [ paper ] [ supplemental ] [ video ] [ code ]
Luxin Zhang, Pascal Germain, Yacine Kessaci, Christophe Biernacki (ECML 2020)

Landmark-based Ensemble Learning with Random Fourier Features and Gradient Boosting [ paper ] [ video ]
Léo Gautheron, Pascal Germain, Amaury Habrard, Guillaume Metzler, Emilie Morvant, Marc Sebban, Valentina Zantedeschi (ECML 2020)

PAC-Bayesian Contrastive Unsupervised Representation Learning [ proceedings ] [ supplemental ] [ arXiv ] [ bibtex ] [ video ] [ code ]
Kento Nozawa, Pascal Germain, Benjamin Guedj (UAI 2020)

PAC-Bayes and Domain Adaptation [ published version ] [ arXiv ] [ bibtex ]
Pascal Germain, Amaury Habrard, François Laviolette, Emilie Morvant (Neurocomputing 2020)

Improved PAC-Bayesian Bounds for Linear Regression [ proceedings ] [ arXiv ] [ bibtex ]
Vera Shalaeva, Alireza Fakhrizadeh Esfahani, Pascal Germain, Mihaly Petreczky (AAAI 2020)

Dichotomize and Generalize: PAC-Bayesian Binary Activated Deep Neural Networks [ proceedings ] [ arXiv ] [ bibtex ] [ video ] [ code ]
Gaël Letarte, Pascal Germain, Benjamin Guedj, François Laviolette (NeurIPS 2019)

Pseudo-Bayesian Learning with Kernel Fourier Transform as Prior [ pdf, supplemental ] [ bibtex ] [ poster ] [ code, datasets ]
Gaël Letarte, Emilie Morvant, Pascal Germain (AISTATS 2019)

Multiview Boosting by Controlling the Diversity and the Accuracy of View-specific Voters [ published version ] [ arXiv preprint ]
Anil Goyal, Emilie Morvant, Pascal Germain, Massih-Reza Amini (Neurocomputing 2019)

Domain-Adversarial Training of Neural Networks [ pdf ] [ bib ] [ source code: shallow version | deep version ] [ data ]
Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, Mario Marchand, Victor Lempitsky (JMLR 2016, and Springer 2017*)
*A slighlty shorter version of the JMLR version is published as a book chapter in Domain Adaptation in Computer Vision Applications (Editor: Gabriela Csurka).

PAC-Bayesian Analysis for a two-step Hierarchical Mutliview Learning Approach [ pdf ]
Anil Goyal, Emilie Morvant, Pascal Germain, Massih-Reza Amini (ECML 2017)

PAC-Bayesian Theory Meets Bayesian Inference [ paper ] [ spotlight: video | slides ] [ poster ] [ code ]
Pascal Germain, Francis Bach, Alexandre Lacoste, Simon Lacoste-Julien (NIPS 2016)

A New PAC-Bayesian Perspective on Domain Adaptation [ pdf ] [ supplemental ] [ bib ] [ source code ] [ data ]
Pascal Germain, Amaury Habrard, François Laviolette, Emilie Morvant (ICML 2016)

PAC-Bayesian Bounds based on the Rényi Divergence [ paper ] [ bib ] [ poster ]
Luc Bégin, Pascal Germain, François Laviolette, Jean-Francis Roy (AISTATS 2016)

Risk Bounds for the Majority Vote: From a PAC-Bayesian Analysis to a Learning Algorithm [ pdf ] [ bib ] [ source code ]
Pascal Germain, Alexandre Lacasse, François Laviolette, Mario Marchand and Jean-Francis Roy (JMLR 2015)

PAC-Bayesian Theory for Transductive Learning [ paper, supplemental ] [ bib ] [ poster ] [ source code ]
Luc Bégin, Pascal Germain, François Laviolette, Jean-Francis Roy (AISTATS 2014)

A PAC-Bayesian Approach for Domain Adaptation with Specialization to Linear Classifiers [ paper, supplemental ] [ bib ] [ source code ] [ data ] [ extended version ]
Pascal Germain, Amaury Habrard, François Laviolette, Emilie Morvant (ICML 2013)

A Pseudo-Boolean Set Covering Machine [ pdf ]
Pascal Germain, Sébastien Giguère, Jean-Francis Roy, Brice Zirakiza, François Laviolette, and Claude-Guy Quimper (CP 2012)

A PAC-Bayes Sample Compression Approach to Kernel Methods [ paper ] [ supplemental ]
Pascal Germain, Alexandre Lacoste, Francois Laviolette, Mario Marchand, and Sara Shanian (ICML 2011)

From PAC-Bayes Bounds to KL Regularization [ pdf ]
Pascal Germain, Alexandre Lacasse, Francois Laviolette, Mario Marchand, and Sara Shanian (NIPS 2009)

PAC-Bayesian Learning of Linear Classifier [ pdf ]
Pascal Germain, Alexandre Lacasse, Francois Laviolette, and Mario Marchand (ICML 2009)

A PAC-Bayes Risk Bound for General Loss Functions [ pdf ]
Pascal Germain, Alexandre Lacasse, Francois Laviolette, and Mario Marchand (NIPS 2006)

PAC-Bayes Bounds for the Risk of the Majority Vote and the Variance of the Gibbs Classifier [ pdf ]
Alexandre Lacasse, Francois Laviolette, Mario Marchand, Pascal Germain, and Nicolas Usunier (NIPS 2006)

Ph.D. Thesis
Généralisations de la théorie PAC-bayésienne pour l’apprentissage inductif, l’apprentissage transductif et l’adaptation de domaine [ pdf (french) ] [ slides (french) ]
Pascal Germain (Université Laval, 2015)

Master's Thesis
Algorithmes d'apprentissage automatique inspirés de la théorie PAC-Bayes [ pdf (french) ] [ bib ] [ english abstract ]
Pascal Germain (Université Laval, 2009)

Selected Talks

27/07/2023 : PAC-Bayesian Learning: A tutorial [ slides ]
PAC-Bayes Meets Interactive Learning Workshop @ ICML 2023 (Hawaii, US)

06/06/2019 : PAC-Bayesian Learning and Neural Networks; The Binary Activated Case [ slides ]
51es Journées de Statistique (Nancy, France)

06/03/2019 : Réseau de neurones artificiels et apprentissage profond [ slides (french) ]
Journée de l'Enseignement de l'Informatique et de l'Algorithmique (Université de Lille, France)

24/01/2017 : Generalization of the PAC-Bayesian Theory, and Applications to Semi-Supervised Learning [ slides ]
Modal Seminar (INRIA Lille, France)

20/06/2016 : A New PAC-Bayesian Perspective on Domain Adaptation [ slides ]
ICML (New-York, US)

02/06/2016 : Variations on the PAC-Bayesian Bound [ slides ]
Bayes in Paris (École nationale de la statistique et de l'administration économique - ENSAE, Paris, France)

31/03/2016 : A Representation Learning Approach for Domain Adaptation [ slides ] [ Proof by Twitter ]
Data Intelligence Group Seminar (Laboratoire Hubert-Curien / Université Jean-Monnet, St-Étienne, France)

01/03/2016 : A Representation Learning Approach for Domain Adaptation [ slides ]
TAO Seminars (INRIA Saclay / CNRS / Université Paris-Sud, Orsay, France)

25/11/2015 : PAC-Bayesian Theory and Domain Adaptation Algorithms [ slides ]
SIERRA Seminars (INRIA Paris / CNRS / ENS, Paris, France)

13/12/2014 : Domain-Adversarial Neural Networks [ slides ] [ workshop paper ]
NIPS 2014 Workshop on Transfer and Multi-task learning: Theory Meets Practice (Montreal, Quebec, Canada)

07/12/2012 : PAC-Bayesian Learning and Domain Adaptation [ slides ]
NIPS 2012 Workshop: Multi-trade-off in Machine Learning (Lake Tahoe, Nevada, US)

05/04/2013 : L'adaptation de domaine en apprentissage automatique: introduction et approche PAC-Bayésienne [ slides (french) ]
Séminaires du département d'informatique et de génie logiciel (Université Laval, Quebec, Canada)

09/10/2012 : A Pseudo-Boolean Set Covering Machine [ slides ]
18th International Conference on Principles and Practice of Constraint Programming (Quebec city, Quebec, Canada)

03/04/2009 : Rudiments de l'apprentissage automatique et de la classification (ainsi que quelques notions plus avancées!) [ slides (french) ]
Séminaires de l'Association des étudiant(e)s gradué(e)s en informatique à Laval (Université Laval, Québec, Canada)

Teaching / Enseignement (in French)