Conférencière invitée
Mercredi 30 juin à 9h
Titre : Aspects of transparency in Reinforcement Learning.
Résumé : While transparency and explainability of machine learning models have recently received quite some attention, this is much less explored in the context of Reinforcement Learning. In this talk I will discuss the different aspects to this problem and provide an overview of the state-of-art, including some of our own work on policy distillation. Finally, I will also briefly touch upon fairness aspects in Reinforcement Learning.
Biographie : Ann Nowé, director of the VUB AI-lab, received her M.S. degree in Mathematics from Universiteit Gent, Belgium, in 1987, and her PhD in AI from Vrije Universiteit Brussels (VUB), Belgium, in collaboration with Queen Mary and Westfield College, University of London, U.K., in 1994. Her major area of interest is machine learning, in particular reinforcement learning, including muti-agent and multi-criteria settings. She currently holds a Francqui research chair to investigate how to make machine learning models more transparent and explainable to users. She is the former chairman of the BNVKI and was a board member of ECCAI. Presently, she is a board member of IFAAMAS, The International Foundation for Autonomous Agents and Multiagent Systems. She was co-PC chair of AAMAS 2021.