The Cross-Entropy Method and Its Application to Inverse Problems

This paper explores the potential of a new evolutionary method-the cross-entropy (CE) method in solving continuous inverse problems. The CE method is a new entrant to the family of metaheuristics, and there are few applications of this method in computational electromagnetics. In the implementation of the cross-entropy method, an adaptive updating formula for the smoothing parameter is proposed to guarantee a satisfactory balance between exploring and exploiting searches. A mutation phase is integrated to prevent the algorithm from being trapped in a local optimal solution, and a different stop criterion is used. The proposed algorithm is evaluated on both low- and high-frequency inverse problems with promising and encouraging results.