Gene Expression Programming Based on Diversity-Guided Grading Evolution
Qiu Jiang-tao · Journal of Sichuan University · 2006
Function mining algorithm based on traditional Gene Expression Programming(GEP) and other improved algorithm may still lead to local optimum trap.To solve this problem,a new algorithm based on Genome Diversity-Guided(DG-GEP) in grading evolution was proposed.The definition of GEP evolution phase and genome diversity evaluation model were given.The concept of anagenesis factor to describe evolution phases and strategy of grading evolution were proposed.The means of dynamic genetic operators and population control were used to make the evolution escape from localization trap quickly.The experiment showed that the new algorithm decreases the generations-stagnancy over 65% and increases the average fitness of colony over 12%.