Chance constrained two-player Zero-Sum Games with a Deep Learning Approach

Abstract: 
In this talk, we present a new deep learning approach for predicting saddle points in chance constrained two-player zero-sum games. Our method combines neurodynamic optimization and deep neural networks. We model the stochastic two-player zero-sum game as an ordinary differential equation (ODE) system using neurodynamic optimization. Then, we develop a neural network to approximate the solution of the ODE system, which includes the saddle point prediction for the game problem. We propose a specialized algorithm for training the neural network to enhance the accuracy of the saddle point prediction. Numerical experiments show the performances of our approach when compared to the state-of-the-art.

Short Biography:
Abdel Lisser is Full Professor at the University Paris Saclay since 2001. He was heading the Graph theory and Combinatorial Optimization group at Computer Science Laboratory from 2006 to 2014. He moved to Signals and Systems Laboratory at CentraleSupelec in 2020. He was heading research group at France Telecom Research Center from 1996 to 2001. He got his Ph.D at the University of Paris Dauphine in Computer Science in 1987 and the Habilitation thesis at the University of Paris Nord in 2000. His main research area is stochastic optimization especially chance constrained optimization together with conic optimization. He also works on chance constrained stochastic games and Markov Decision Processes with chance constraints. Recently, he started working on dynamical neural networks and applications of machine learning in stochastic optimization. Applications areas are mainly energy planning optimization, autonomous vehicles and network design problems. He has already run different national and international projects in the recent years.