TY - GEN
T1 - Online Performative Gradient Descent for Learning Nash Equilibria in Decision-Dependent Games
AU - Zhu, Zihan
AU - Fang, Ethan X.
AU - Yang, Zhuoran
N1 - Publisher Copyright:
© 2023 Neural information processing systems foundation. All rights reserved.
PY - 2023
Y1 - 2023
N2 - We study multi-agent games within the innovative framework of decision-dependent games, which establishes a feedback mechanism that population data reacts to agents' actions and further characterizes the strategic interactions among agents. We focus on finding the Nash equilibrium of decision-dependent games in the bandit feedback setting. However, since agents are strategically coupled, classical gradient-based methods are infeasible without the gradient oracle. To overcome this challenge, we model the strategic interactions by a general parametric model and propose a novel online algorithm, Online Performative Gradient Descent (OPGD), which leverages the ideas of online stochastic approximation and projected gradient descent to learn the Nash equilibrium in the context of function approximation for the unknown gradient. In particular, under mild assumptions on the function classes defined in the parametric model, we prove that the OPGD algorithm finds the Nash equilibrium efficiently for strongly monotone decision-dependent games. Synthetic numerical experiments validate our theory.
AB - We study multi-agent games within the innovative framework of decision-dependent games, which establishes a feedback mechanism that population data reacts to agents' actions and further characterizes the strategic interactions among agents. We focus on finding the Nash equilibrium of decision-dependent games in the bandit feedback setting. However, since agents are strategically coupled, classical gradient-based methods are infeasible without the gradient oracle. To overcome this challenge, we model the strategic interactions by a general parametric model and propose a novel online algorithm, Online Performative Gradient Descent (OPGD), which leverages the ideas of online stochastic approximation and projected gradient descent to learn the Nash equilibrium in the context of function approximation for the unknown gradient. In particular, under mild assumptions on the function classes defined in the parametric model, we prove that the OPGD algorithm finds the Nash equilibrium efficiently for strongly monotone decision-dependent games. Synthetic numerical experiments validate our theory.
UR - http://www.scopus.com/inward/record.url?scp=85196400982&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85196400982&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85196400982
T3 - Advances in Neural Information Processing Systems
BT - Advances in Neural Information Processing Systems 36 - 37th Conference on Neural Information Processing Systems, NeurIPS 2023
A2 - Oh, A.
A2 - Neumann, T.
A2 - Globerson, A.
A2 - Saenko, K.
A2 - Hardt, M.
A2 - Levine, S.
PB - Neural information processing systems foundation
T2 - 37th Conference on Neural Information Processing Systems, NeurIPS 2023
Y2 - 10 December 2023 through 16 December 2023
ER -