Biography

I am a PhD student in Machine Learning at CNRS in the Inria ScooL team (formerly SequeL), under the direction of Emilie Kaufmann and Odalric-Ambrym Maillard.

Motivated by practical problems, I explore alternatives to the classical UCB and Thompson Sampling algorithms in Multi-Armed Bandits. I mainly explored two families of algorithms. The first one includes sub-sampling approaches (SDA), allowing for “greedy” comparisons between arms with strong theoretical guarantees in the MAB and non-stationary MAB settings. The second one is a generalization of the Thompson Sampling for bounded distribution, oriented towards practicioners by considering risk-awareness and the question of the sensitivity to model misspecification.

I am also generally interested in reinforcement learning, statistics, and machine learning.