Co-evolutionary self-adaptive Differential Evolution with a uniform-distribution update rule

Differential Evolution (DE) is a simple evolutionary algorithm which is inherently adaptive. This is due to the fact that the mutation amount is derived from the difference of randomly chosen members of the population, which is automatically reduced as the population diversity drops. The process is however governed by an important weighing parameter F, to which the global properties of the DE are very sensitive. Large F can lead to significant reductions in convergence speed, whilst small F can cause the algorithm to get stuck. In this paper, a simple co-evolutionary process is proposed to automatically update the F parameter during the optimization process based on a uniformly distributed update rule. The behavior of the adaptive DE is studied and investigated with some benchmark functions.