Professor Patrick Rebeschini

Professor of Statistics and Machine Learning

Biographical Sketch

I have a Ph.D. in Operations Research and Financial Engineering from Princeton University (2014). After that, I joined the Yale Institute for Network Science at Yale University. I worked two years as a Postdoctoral Associate in the Electrical Engineering Department, and one year as an Associate Research Scientist with a joint appointment as a Lecturer in the Computer Science Department at Yale.

Research Interests

My research interests lie at the intersection of probability, statistics, and computer science. I am interested in the investigation of fundamental principles in high-dimensional probability, statistics and optimisation to design computationally efficient and statistically optimal algorithms for machine learning.

Publications

Huh, J. and Rebeschini, P. (2024) “Generalization bounds for label noise stochastic gradient descent”, in Proceedings of the 27th International Conference on Artificial Intelligence and Statistics (AISTATS 2024). Journal of Machine Learning Research, pp. 1360–1368.

Alfano, C., Towers, S., Sapora, S., Lu, C. and Rebeschini, P. (2024) “Meta-learning the mirror map in policy mirror descent”, arXiv.

Johnson, E., Pike-Burke, C. and Rebeschini, P. (2024) “Sample-efficiency in multi-batch reinforcement learning: the need for dimension-dependent adaptivity”, in Proceedings of the International Conference on Learning Representations (ICLR 2024). OpenReview.

Johnson, E., Pike-Burke, C. and Rebeschini, P. (2023) “Optimal convergence rate for exact policy mirror descent in discounted Markov decision processes”, in Advances in Neural Information Processing Systems. NeurIPS, pp. 76496–76524.

Kanade, V., Rebeschini, P. and Vaskevicius, T. (2023) “The statistical complexity of early-stopped mirror descent”, Information and Inference: A Journal of the IMA, 12(4), pp. 3010–3041.

Huh, J. and Rebeschini, P. (2023) “Generalization bounds for label noise stochastic gradient descent”, arXiv.

Johnson, E., Pike-Burke, C. and Rebeschini, P. (2023) “Sample-Efficiency in Multi-Batch Reinforcement Learning: The Need for Dimension-Dependent Adaptivity”, arXiv.

Johnson, E., Pike-Burke, C. and Rebeschini, P. (2023) “Optimal Convergence Rate for Exact Policy Mirror Descent in Discounted Markov Decision Processes”, arXiv.

Alfano, C., Yuan, R. and Rebeschini, P. (2023) “A Novel Framework for Policy Mirror Descent with General Parameterization and Linear Convergence”, arXiv.

Wu, F. and Rebeschini, P. (2022) “Nearly minimax-optimal rates for noisy sparse phase retrieval via early-stopped mirror descent”, Information and Inference: a Journal of the IMA, 12(2), pp. 633–713.

Contact Details

College affiliation: Tutorial Fellow at University College

Email: patrick.rebeschini@stats.ox.ac.uk

Web: https://www.stats.ox.ac.uk/~rebeschi/

Research Groups

Statistical Theory and Methodology

Computational Statistics and Machine Learning