A Two-Timescale Neurodynamic Approach to Minimax Portfolio Optimization

Abstract

This paper is concerned with asset allocation based on two-timescale neurodynamic optimization. The portfolio optimization in classical mean-variance framework is reformulated as a minimax portfolio selection problem and a two-timescale neurodynamic approach is developed to solve the problem. The neurodynamic approach incorporates a recurrent neural network (RNN) operating on two different timescales. Computational results show the efficacy and performance of the developed approach to asset allocation.