A Mixed Strategies Differential Evolution Based on Fitness Landscapes Features

Abstract

Differential Evolution (DE) is obviously one of the most powerful and versatile evolutionary algorithm for solving many complex real-world optimisation problems e ciently in recent times. Since their introduction in 1995, the focus of research in DEs has largely been on the variants side with so many new algorithms proposed on the original DE algorithm. However, each operator is only suitable for certain tness landscapes, therefore some types of optimization problems that cannot be solved efficiently. To help in this task, this paper presents a new mixed strategies DE based on fitness landscapes, named as FLDE, the optimal variation strategy is selected by extracting the local fitness landscape features of each generation population and combing the probability distributions of the unimodal and the multi-modal in the local fitness landscape. The proposed algorithm is tested on a suite of 13 benchmark functions, the experimental results demonstrated the advantages of this work in high search dimension that it could ensure the population move to the better fitness landscape and then speed up the convergence to global optimal as well as avoid falling into the local optima.