Solving Nash Equilibrium Scalably via Deep-Learning-Augmented Iterative Algorithms
Keywords: Nash Equilibrium, Game Theory, Deep Learning
TL;DR: We propose a deep-learning augmented iterative solver for Nash equilibrium, effectively avoiding the curse of dimensionality by reducing the time complexity to a polynomial level.
Abstract: Computing the Nash Equilibrium (NE) is a fundamental yet computationally challenging problem in game theory. Although recent approaches have incorporated deep learning techniques to tackle this intractability, most of them still struggle with scalability when the number of players increases, due to the exponential growth of computational cost. Inspired by the efficiency of classical learning dynamics methods, we propose a deep learning-augmented Nash equilibrium solver, named Deep Iterative Nash Equilibrium Solver (DINES), based on a novel framework that integrates deep learning into iterative algorithms to solve Nash Equilibria more efficiently. Our approach effectively reduces time complexity to a polynomial level and mitigates the curse of dimensionality by leveraging query-based access to utility functions rather than requiring the full utility matrix. Experimental results demonstrate that our approach achieves better or comparable approximation accuracy compared to existing methods, while significantly reducing computational expense. This advantage is highlighted in large-scale sparse games, which is previously intractable for most existing deep-learning-based methods.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13654
Loading