Improved Differential Evolution with Parameter Adaption Based on Population Diversity
Cheng Hongtan, Liu Zhaoguang · 2018
The differential evolution algorithm is an important branch of the bionic intelligent computation, which uses the Darwinian population's evolutionary principle: survival of the fittest and survival of the fittest. Due to the simple implement and few parameters, many researchers have invested into the study of the algorithm and proposed a large number of differential evolution variants. For the existing differential evolution algorithm, once the size of the population is determined, the size of the search range is fixed. Based on the global diversity of population, we focus on controlling the value of the search parameters p. In the proposal, after normalizing the population diversity, each individual will select its unique search scope according to the diversity conditions. Therefore, the proposed method can balance between the global search and the local search. According to our extensive experimental results on various benchmark functions, the proposed method outperform other compared advanced algorithms.