An Improved Differential Evolution Algorithm and Its Application to Large-Scale Artificial Neural Networks

Abstract

A new differential evolution (DE) algorithm is presented in this paper. The proposed algorithm monitors the evolutionary progress of each individual and assigns appropriate control parameters depends on whether the individual is successfully evolved or not. We conducted the performance evaluation on CEC 2014 benchmark problems and confirmed that the proposed algorithm outperformed than the conventional DE algorithm. In addition, we apply the proposed DE algorithm as an optimization technique of training large scale multilayer perceptron. We conducted the performance evaluation on an artificial neural network that has approximately 1,000 weights and confirmed again that the proposed algorithm performed better than the conventional DE algorithm. As a result, we proposed a new DE algorithm that has better optimization performance for solving large-scale global optimization problems.