The bandit problem is a simplified instance of a reinforcement learning problem where an algorithm learns the optimal action by interacting with an environment. This talk will focus more specifically on structured bandits, in which a notion of similarity between different actions can be exploited to speed-up learning. This variant of the problem will be presented through an application to super-resolution imaging parameters optimization.

The presentation will be given in French and the slides will be in English.

The seminar will be presented at 11:30 a.m. in room PLT-3370.