A Niching Gene Expression Programming Algorithm Based on Parallel Model

GEP is a biologically motivated machine learning technique used to solve complex multitude problems. Similar to other evolution algorithms, GEP is slow when dealing with a large number of population. Considering that the parallel GEP has great efficiency and the niching method can keep diversity in the process of exploring evolution, a niching GEP algorithm based on parallel model is presented and discussed in this paper. In this algorithm, dividing the population to the niche nodes in sub-populations can solves the same problem in less computation time than it would take on a single process. Experimental results on sequence induction, function finding and sunspot prediction demonstrate its advantages and show that the proposed method takes less computation time but with higher accuracy.