An improved multi-objective cross-entropy method for inverse problems

Abstract

An improved vector cross-entropy (CE) method is proposed to provide a potential candidate for solving multi-objective inverse problems with a large number of variables. To balance the exploitation and exploration searches, the whole iterative process is divided into two phases: diversification and intensification phases. Different parameter evolutionary mechanisms of the probability density functions (pdfs) are proposed for different phases. To speed up the convergence rate, a dynamic evolutionary mechanism is proposed. To enhance the diversity of the sampling points, a mutation manipulation is introduced. The ZDT test functions and a high frequency inverse problem are used as the case studies to testify the effectiveness and efficiency of the proposed method.