No próximo dia 11 de junho de 2021 pelas 11h00, Sébastien Gerchinovitz irá dar uma palestra intitulada "Online learning: a quick tour of bandit algorithms".
A palestra é organizada pelo DCC-FCUP e pelo grupo de investigação LIAAD-INESC TEC e é aberta a todos os interessados.
A sessão será de forma remota. A participação é gratuita mas a inscrição é obrigatória.
Sébastien Gerchinovitz is a research scientist at IRT Saint Exupéry Toulouse, France, working in the DEEL project on machine learning theory and applications to safety-critical systems. Sébastien is also associate researcher at the Institut de Mathématiques de Toulouse, and a member of the Game Theory and Artificial Intelligence chair within the Toulouse AI Institute (ANITI). He is currently on leave from Université Toulouse III - Paul Sabatier where he holds an assistant professor position. Sébastien received a PhD in Mathematics at Ecole Normale Supérieure, Paris. His main research topics are online learning, statistical learning theory, and deep learning theory. He has been in the Program Committees of ALT 2018 & 2019, and COLT 2020 & 2021. He has had the opportunity to do several research visits at University of Milan, UC Berkeley, and CWI Amsterdam.
Online learning: a quick tour of bandit algorithms
Online learning is a family of repeated decision-making problems where a learner interacts with an environment by choosing actions and receiving rewards round after round, with the goal, e.g., to maximize the total reward. Bandit algorithms address the special case where the obtained rewards are the only feedback given to the learner. This covers applications such as clinical trials, recommender systems, or ad auction optimization. In this talk, we will present a quick tour of bandit algorithms, with different learning-theoretic bounds in different settings. We will also discuss several applications and some connections with global optimization problems.