Elvis Dohmatob, PhD

Associate Professor, Computer Science and Software Engineering

Research areas: artificial intelligence, machine learning, learning theory, algorithms, neural networks, neural scaling laws, adversarial robustness, algorithmic bias, optimization, determinantal point processes

Contact information

Email:

elvis.dohmatob@concordia.ca

Biography

I am a researcher working on various topics in artificial intelligence (AI) and machine learning (ML), with a theoretical flavor. I joined Concordia as a professor in 2024. I'm also an affiliate faculty at the Mila Institute. Prior to Concordia, I held positions at INRIA (Paris, France), Criteo (Paris, France), and FAIR/Meta (Paris, France).

My current research agenda focuses around the following themes:

Trustworthy AI/ML (algorithmic bias, adversarial robustness, out-of-distribution generalization, etc.),
Learning Theory (neural scaling laws, model collapse, etc.),
Neural Networks (attention, associative memories, etc.),
Representation Learning,
Optimization (convex, discrete, etc.)

Publications

2024

E. Dohmatob, Y. Feng, P. Yang, F. Charton, J. Kempe, "A Tale of Tails: Model Collapse as a Change of Scaling Laws", In International Conference on Machine Learning (ICML), volume 235 of Proceedings of Machine Learning Research, 2024
E. Dohmatob, Y. Feng, J. Kempe, "Model Collapse Demystified: The Case of Regression", In Advances in Neural Information Processing Systems (NeurIPS), volume 37, 2024
V. Cabannes, E. Dohmatob, A. Bietti, "Scaling Laws for Associative Memories", In The Twelfth International Conference on Learning Representations (ICLR), 2024
E. Dohmatob, M. Scetbon, "Precise Accuracy / Robustness Tradeoffs in Regression: Case of General Norms", In International Conference on Machine Learning (ICML), volume 235 of Proceedings of Machine Learning Research, 2024
E. Dohmatob, "Consistent Adversarially Robust Linear Classification: Non-Parametric Setting", In International Conference on Machine Learning (ICML), volume 235 of Proceedings of Machine Learning Research, 2024
E. Dohmatob, Y. Feng, A. Subramonian, J. Kempe, "Strong Model Collapse", ArXiv preprint
A. Subramonian, S. Bell, L. Sagun, E. Dohmatob, "An Effective Theory of Bias Amplification", ArXiv preprint
Y. Feng, E. Dohmatob, P. Yang, F. Charton, J. Kempe, "Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement", ArXiv preprint