COMARL Virtual Seminars – Michael Bowling, Hindsight Rationality: Alternatives to Nash

On January 28, Fellow and Canada CIFAR AI Chair at Amii Michael Bowling (Professor of Computing Science at the University of Alberta) presented "Hindsight Rationality: Alternatives to Nash" as the inaugural talk in the COMARL Seminar Series.

The COMARL Seminar Series explores Challenges and Opportunities for Multi-Agent Reinforcement Learning with biweekly seminars hosted virtually by COMARL seminar organizers. The seminars are organized by a team of researchers from DeepMind, Delft University of Technology and Northeastern University. These seminars highlight researchers who share their work in multiagent reinforcement learning and game theory.

Bowling discusses the contents of two papers, co-authored by several researchers at Amii: Hindsight and Sequential Rationality of Correlated Play and Efficient Deviation Types and Learning for Hindsight Rationality in Extensive-Form Games.

In his talk, he discusses some of the often unstated principles common in multiagent learning research (and emergent communication work), suggesting that these unstated principles may be responsible for holding us back. In response, the talk offers an alternative set of principles, which leads to the view of hindsight rationality with connections to online learning and correlated equilibria. Bowling then describes some recent technical work that seeks to understand how we can build increasingly powerful algorithms for hindsight rationality in sequential decision-making settings.

Watch the full presentation below:

Learn how Amii advances world-leading artificial intelligence and machine learning research: visit our Research page.