One parameter differential evolution (OPDE) for numerical benchmark problems

Abstract

Differential Evolution (DE) can be simplified in the sense that the number of existing parameter is decreased from two parameters to only one parameter. We eliminate the scaling factor, F, and replace this by a uniform random number within [0, 1]. As such, it is easy to tune the crossover rate, CR, through parameter sensitivity analysis. In this analysis, the algorithm is run for 50 trials from 0.1 to 1.0 with a step increment of 0.1 on 23 benchmark problems. Results show that using the optimal CR, there is room for improvement in some of the benchmark problems. With the advantage and simplicity of a single parameter, it is significantly easier to tune this parameter and thus take the full advantage of the algorithm. The proposed algorithm here has a significant benefit when applied to real-world problems as it saves time in obtaining the best parameter setting for optimal performance. © 2013 Springer-Verlag Berlin Heidelberg.