• generative adversarial networks
• mean field games

We analyse the connection between Mean Field Games (MFGs) and a popular Machine Learning model, namely Generative Adversarial Networks (GANs). From the game theoretical perspective, GANs can be interpreted as MFGs under Pareto Optimality condition. On the other side, we take advantage of the adversarial nature of GANs in order to numerically approximate Mean Field Equilibria, which are expressed as solutions of a system of coupled PDEs: a forward Fokker-Plank equation and a backward Hamilton-Jacobi-Bellman equation.