An empirical comparison of differential evolution variants for high dimensional function optimization

In this paper, we present an empirical comparison of some differential evolution (DE) variants to solve high dimensional optimization problems. The aim is to identify the behavior and scalability of DE variants. Most studies on DE are obtained using low-dimensional problems (smaller than 100) , which are relatively small for many real-world problems. We have chosen four problems grouped by feature: unimodal and separable, unimodal and nonseparable, multimodal and separable, and multimodal and nonseparable. Fourteen variants were implemented and tested on four benchmark problems for dimensions of 30, 100, 500 and 1000. The value for the parameter CR is decided based on a bootstrap test conducted for 30 dimensions, and the same CR value is adopted for the dimensions 100, 500 and 1000 also. The analysis is done based on the results obtained for 100 runs, for each variant-function-dimension combination.