MULTI-AGENT PORTFOLIO SELECTION AND DEEP LEARNING APPLICATIONS

Abstract

This thesis consists of three parts. In part I, we examine the investor's optimal choice under information frictions. The deep learning method is applied to solve the optimal portfolio of the investors. In part I, we solve a mean-field game with stochastic return coefficient. The model is motivated by the partial information game in part I and we emphasize on the theoretical contribution of our results. In part III, we design several deep neural algorithms to solve portfolio selection problems. This part contributes mainly as computation methods since the problems proposed are highly non-trivial in the classical sense.