Solving Neural Min-Max Games: The Role of Architecture, Initialization & Dynamics

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Patel, Deep, Vlatakis-Gkaragkounis, Emmanouil-Vasileios
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908681475457024
author Patel, Deep
Vlatakis-Gkaragkounis, Emmanouil-Vasileios
author_facet Patel, Deep
Vlatakis-Gkaragkounis, Emmanouil-Vasileios
contents Many emerging applications - such as adversarial training, AI alignment, and robust optimization - can be framed as zero-sum games between neural nets, with von Neumann-Nash equilibria (NE) capturing the desirable system behavior. While such games often involve non-convex non-concave objectives, empirical evidence shows that simple gradient methods frequently converge, suggesting a hidden geometric structure. In this paper, we provide a theoretical framework that explains this phenomenon through the lens of hidden convexity and overparameterization. We identify sufficient conditions - spanning initialization, training dynamics, and network width - that guarantee global convergence to a NE in a broad class of non-convex min-max games. To our knowledge, this is the first such result for games that involve two-layer neural networks. Technically, our approach is twofold: (a) we derive a novel path-length bound for the alternating gradient descent-ascent scheme in min-max games; and (b) we show that the reduction from a hidden convex-concave geometry to two-sided Polyak-Łojasiewicz (PŁ) min-max condition hold with high probability under overparameterization, using tools from random matrix theory.
format Preprint
id arxiv_https___arxiv_org_abs_2512_00389
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Solving Neural Min-Max Games: The Role of Architecture, Initialization & Dynamics
Patel, Deep
Vlatakis-Gkaragkounis, Emmanouil-Vasileios
Machine Learning
Computer Science and Game Theory
Many emerging applications - such as adversarial training, AI alignment, and robust optimization - can be framed as zero-sum games between neural nets, with von Neumann-Nash equilibria (NE) capturing the desirable system behavior. While such games often involve non-convex non-concave objectives, empirical evidence shows that simple gradient methods frequently converge, suggesting a hidden geometric structure. In this paper, we provide a theoretical framework that explains this phenomenon through the lens of hidden convexity and overparameterization. We identify sufficient conditions - spanning initialization, training dynamics, and network width - that guarantee global convergence to a NE in a broad class of non-convex min-max games. To our knowledge, this is the first such result for games that involve two-layer neural networks. Technically, our approach is twofold: (a) we derive a novel path-length bound for the alternating gradient descent-ascent scheme in min-max games; and (b) we show that the reduction from a hidden convex-concave geometry to two-sided Polyak-Łojasiewicz (PŁ) min-max condition hold with high probability under overparameterization, using tools from random matrix theory.
title Solving Neural Min-Max Games: The Role of Architecture, Initialization & Dynamics
topic Machine Learning
Computer Science and Game Theory
url https://arxiv.org/abs/2512.00389