Dynamic Fitness Landscape Analysis on Differential Evolution Algorithm

Abstract

Dynamic fitness landscape analyses mainly try to figure out the performance of evolutionary algorithms through some simple graphs and effective data. In this paper, we focus on one of evolutionary algorithms named as differential evolution (DE) algorithm. Six benchmark functions we selected because of different properties are involved in our experiments using metrics of dynamic fitness landscape analyses to test. According to experimental results, they shows obviously that differential evolution algorithm can calculate low dimension of benchmark functions and is very hard to handle high dimension. When a benchmark function becomes more and more complicate within higher dimension, sometimes differential evolution algorithm can get good results, but most of time there is no result at all. Dynamic fitness landscape analyses truly obtain experimental results and more details as differential evolution algorithm.