Learning equilibria in games with zeroth-order information
Event details
| Date | 16.05.2023 |
| Hour | 16:15 › 17:15 |
| Speaker | Maryam Kamgarpour (EPFL) |
| Location | |
| Category | Conferences - Seminars |
| Event Language | English |
I address the question of learning equilibria in convex and non-convex games under zeroth-order information. In the convex setting, I will present our learning algorithm for monotone games that leverages estimation of game pseudo-gradient. I discuss extensions of the approach to a broader class of games as well as its convergence rates. In the non-convex setting, I will present our no-regret algorithm that leverages probabilistic estimation of a players' cost function. I will compare the regret rate of the algorithm with the state-of-the-art and discuss our extensions to multi agent reinforcement learning.
Practical information
- General public
- Free
- This event is internal
Organizer
- Prof. Fabio Nobile
Contact
- Fabio Nobile, Séverine Eggli