Chargement...
 
Tao

Seminar26042016

April 26th

14:30, R2014 Digiteo Shannon (660) (see location):


Marc Bellemare (Google DeepMind)


Title: Eight Years of Research with the Atari 2600


Abstract:

In this talk I will review some recent and not-so-recent research developments with
the Arcade Learning Environment (ALE), a reinforcement learning interface to the Atari 2600 video
game console. I will discuss some of the core AI advances that have come from research with the
platform, present some surprising discoveries made along the way, and highlight the research
challenges still posed by the ALE. Returning to the ALE's original role as an evaluation platform
for reinforcement learning, I'll describe some of the evaluation challenges posed by using a set
of heterogeneous domains, and review the solutions proposed so far. In the second part of the talk,
I will present our recent work on developing new Bellman-like operators for reinforcement learning,
and their application, of course, to Atari 2600 games.


Contact: odalricambrym.maillard at inria.fr, guillaume.charpiat at inria.fr

Collaborateur(s) de cette page: guillaume .
Page dernièrement modifiée le Lundi 25 avril 2016 15:35:02 CEST par guillaume.