Two-player Games in the Era of Machine Learning

Par

Gauthier Gidel

mila / Université de Montréal
https://gauthiergidel.github.io/

Pour assister au colloque sur Bluejeans, merci de remplir ce formulaire :

https://forms.gle/NB5KfbTpf19ukZ3x6

Résumé:

Adversarial training, a special case of multiobjective optimization, is an increasingly useful tool in machine learning. For example, two-player zero-sum games are important for generative modeling (GANs) and for mastering games like Go or Poker via self-play. A classic result in Game Theory states that one must mix strategies, as pure equilibria may not exist. Surprisingly, machine learning practitioners typically train a single pair of agents – instead of a pair of mixtures – going against Nash’s principle. In this talk I will put learning in two-player zero-sum games on a firm theoretical foundation. I will first introduce a notion of limited-capacity-equilibrium for games played with neural networks for which, as capacity grows, optimal agents – not mixtures – can learn increasingly expressive and realistic behaviors. I will then focus on the training of such games by countering some common misconceptions about the difficulties of two-player games optimization and proposing to extend techniques designed for variational inequalities to the training of GANs.

Biographie :

Gauthier Gidel is a PhD candidate at Université de Montréal as a member of Mila and DIRO and has been a research intern at ElementAI and DeepMind. His research is a the interface between machine learning, game theory and optimization on which he co-organized NeurIPS workshops in 2018 and 2019. Gauthier is a recipient of the Borealis AI graduate fellowship as well as the DIRO excellence grant.