Short Bio

I am a machine learning researcher with a background in actuarial sciences and a Ph.D. in computer science from University Laval (2026), supervised by Prof. Pascal Germain. My thesis, "Trustworthy Machine Learning through Simplicity: Interpretability, Explainability, and Generalization Guarantees", was defended with honors (mention «Excellence»). I am now looking for a postdoc opportunity to expand my contributions in the fields of PAC-Bayesian (machine learning) theory and interpretability/explainability in machine learning.

Research Interests

      
  • Machine learning theory
  • Generalization guarantees
  • Deep learning
                               
  • Interpretability
  • Explainability
  • Ethics

Publications

Peer-Reviewed Works (conferences main track, journals)

Generalization Bounds via Meta-Learned Model Representations: PAC-Bayes and Sample Compression Hypernetworks [publication]
Benjamin Leblanc, Mathieu Bazinet, Nathaniel D'Amours, Alexandre Drouin, Pascal Germain (ICML, 2025)
Seeking Interpretability and Explainability in Binary Activated Neural Networks [publication] [preprint]
Benjamin Leblanc, Pascal Germain (XAI World Conference, 2024)
Application of machine learning tools to study the synergistic impact of physicochemical properties of peptides and filtration membranes on peptide migration during electrodialysis with filtration membranes [publication]
Zain Sanchez-Reinoso, Mathieu Bazinet, Benjamin Leblanc, Jean-Pierre Clément, Pascal Germain, Laurent Bazinet (Journal of Separation and Purification Technology, 2024)
PAC-Bayesian Learning of Aggregated Binary Activated Neural Networks with Probabilities over Representations [publication]
Louis Fortier-Dubois, Benjamin Leblanc, Gaël Letarte, François Laviolette, Pascal Germain (Canadian AI, 2023)

Peer-Reviewed Works (workshops)

Simplicity Suffices for Parameter Noise Injection in Stochastic Gradient Descent [preprint]
Benjamin Leblanc, Louis-Jacob Lebel, Teddy Kana, Richard Kamel (Data Science Meets Optimisation (DSO) Workshop at IJCAI, 2026)
Sample Compression Hypernetworks: From Generalization Bounds to Meta-Learning [publication]
Benjamin Leblanc, Mathieu Bazinet, Nathaniel D'Amours, Alexandre Drouin, Pascal Germain (Neural Compression Workshop at NeurIPS, 2024)
Seeking Interpretability and Explainability in Binary Activated Neural Networks [publication]
Benjamin Leblanc, Pascal Germain (Workshop on Explainable Artificial Intelligence at IJCAI, 2023)
A Greedy Algorithm For Building Compact Binary Activated Neural Networks [publication]
Benjamin Leblanc, Pascal Germain (Montreal AI Symposium, 2022)
Learning Aggregation of Binary Activated Neural Networks with Probabilities over Representations [publication]
Louis Fortier-Dubois, Benjamin Leblanc, Gaël Letarte, François Laviolette, Pascal Germain (Montreal AI Symposium, 2021)

Selected Reports

On the disintegration of the stochastic majority vote: From PAC-Bayesian bounds to a self-bounding algorithm [ArXiv]
Julien Bastian, Benjamin Leblanc, Pascal Germain, Amaury Habrard, Guillaume Metzler, Emilie Morvant, Paul Viallard (2026)
PAC-Bayesian Generalization Guarantees for Fairness on Stochastic and Deterministic Classifiers [ArXiv]
Julien Bastian, Benjamin Leblanc, Pascal Germain, Amaury Habrard, Christine Largeron, Guillaume Metzler, Emilie Morvant, Paul Viallard (2026)
A Framework for Bounding Deterministic Risk with PAC-Bayes: Applications to Majority Votes [ArXiv]
Benjamin Leblanc, Pascal Germain (2025)
On the Relationship Between Interpretability and Explainability in Machine Learning [ArXiv]
Benjamin Leblanc, Pascal Germain (2023)

Mentoring

Louis-Jacob Lebel (Bachelor’s, University Laval, 2025-2026)
Richard Kamel (Bachelor’s, University Laval, 2025-2026)
Teddy Kana (Bachelor’s, University Laval, 2025-2026)

Teaching

Mathématiques pour informaticiens (2023, 2024) - Université Laval, Département d'informatique et de génie logiciel

Affiliations