Research On The Fusion Method And Architecture Of MEP And GEP

Multi-expression programming(MEP)and gene expression programming(GEP)are two important variants of genetic programming(GP).In this paper,based on the existing research work of scholars,we find that the expression generated by operation of MEP and GEP belongs to the category of legal expressions on the terminal set.Therefore,we formed a fusion method MMGEP(MEP mix GEP)by integrating GEP chromosome into MEP gene and adding a position pointer to GEP.This method not only unifies the algorithm framework of MEP and GEP,but also has advantages over single MEP or GEP in performance.Our main work is as follows:1)Some examples of MMGEP changing into GEP and MEP in the framework of MMGEP algorithm are given,and a new coding method of MEP and GEP based on the framework is proposed,which proves the universality of the method in the characteristics of unified framework.2)The idea of sharing evaluation is used to optimize the evolution time of MEP,and the method of symbolic regression is used to prove that MMGEP is superior to single MEP or GEP in accuracy,precision and other basic GP evaluation indexes.3)Summarizes the work done in this paper and gives some suggestions on MMGEP optimization.