Xinying Zou
PhD student at Centre Inria d'Université Côte d'Azur
Dec. 2022 - present, under the supervision of Samir M. Perlaza, Eitan Altman, and Iñaki Esnaola.
PhD student at Centre Inria d'Université Côte d'Azur
Dec. 2022 - present, under the supervision of Samir M. Perlaza, Eitan Altman, and Iñaki Esnaola.
Research Overview:
I am interested in the generalization of machine learning algorithms. My Ph.D. thesis provides a rigorous information-theoretic framework to characterize and improve the generalization capabilities of supervised learning algorithms.
Through distributionally robust optimization with Kullback-Leibler divergence constraints, closed-form expressions for generalization error, model sensitivity, etc., are provided in terms of information measures. Building on a training-dependent minimax framework, an algorithm robustification method is proposed to improve the performance of a given baseline learning algorithm w.r.t. unseen data, which addresses the gap between empirical training performance and reliability under distribution shifts.
Information Theory and Statistical Learning.