Game on Random Environment, Mean-field Langevin System and Neural Networks - Archive ouverte HAL Access content directly
Preprints, Working Papers, ... Year : 2022

Game on Random Environment, Mean-field Langevin System and Neural Networks

(1) , (2) , (3)
1
2
3

Abstract

In this paper we study a type of games regularized by the relative entropy, where the players' strategies are coupled through a random environment variable. Besides the existence and the uniqueness of equilibria of such games, we prove that the marginal laws of the corresponding mean-field Langevin systems can converge towards the games' equilibria in different settings. As applications, the dynamic games can be treated as games on a random environment when one treats the time horizon as the environment. In practice, our results can be applied to analysing the stochastic gradient descent algorithm for deep neural networks in the context of supervised learning as well as for the generative adversarial networks.
Fichier principal
Vignette du fichier
MFL_Game_arxiv.pdf (439.58 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02532096 , version 1 (04-04-2020)
hal-02532096 , version 2 (15-04-2020)

Identifiers

  • HAL Id : hal-02532096 , version 2

Cite

Giovanni Conforti, Anna Kazeykina, Zhenjie Ren. Game on Random Environment, Mean-field Langevin System and Neural Networks. 2020. ⟨hal-02532096v2⟩
300 View
50 Download

Share

Gmail Facebook Twitter LinkedIn More