{"id":"arxiv/2604.05685","url":"https://arxiv.org/abs/2604.05685","tags":["math.NA","math.OC"],"score":0,"title":"Discrete Mean Field Games on Finite Graphs as Initial Value Optimization","domain":"math.NA","source":"arxiv","authors":["Yaxin Feng","Yang Xiang","Haomin Zhou"],"reviews":0,"subject":"math.NA","summary":"In this paper, we propose an initial value fomulation of the discrete mean field games on finite graphs (Graph MFG), and design a neural network based approach to solve it. Graph MFG describes infinite, non-cooperative and interactive homogeneous agents move on node states through the edges to optimize their own goals. Nash Equilibrium of the Graph MFG is characterized by a coupled ordinary differential equations (ODE) system, including the discrete forward continuity equation and the discrete backward Hamilton-Jacobi equation. In this paper, we mainly focus on the potential mean field games…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-04-07","reactions":{}}