The Research On GEP Based On The Open Read Frame

GEP(Gene Expression Programming) adopted the simple linear encoding method which used in GA(Genetic Algorithm) and mapped the linear gene into a complex tree structure by the decoding algorithms. In this way, GEP combined the characteristics of simple individual representing complex problem spaces used in GP(Genetic Programming) and the simple linear encoding method used in GA, which improved the search performance by 2~4 orders of magnitude.However, it has the disadvantages of lower local and global searching efficiency. Aiming to solve the problem, this paper mainly proposed two kinds of improved GEP algorithm:(1)There is non-coding region in the gene through the study of GEP genetic operator and ORF. If the gene recombination or insert sequence occurs in the non-coding region, then the expression trees corresponding to the chromosome will remain unchanged, which will resulting in repeated search in the problem space and reducing the searching efficiency of algorithm. According to defects of the traditional GEP genetic operator, this paper redesigned the recombination and insert sequence operator to improve the searching ability. This kind of operator selects action points from the gene coding region in order to ensure the genetic operation will change the coding region of the gene fragments, so that the gene can be mapped into different tree. The experimental results show that the improved GEP has shortened the evolution and improved the success rate.(2)According to the process of GEP algorithm, the population tends to be the same during the late of the evolution which reflects the week of the global searching capacity. Aiming to solve this problem, this paper proposed a new GEP algorithm which controls the diversity of the population through the proportion of gene decoding structure. This new algorithm combines the population diversification strategy based on gene decoding structure under the ORF, which prevented the evolution into the local convergence. The experimental results show that the improved GEP algorithm can control the diversity of chromosome effectively and prevent the evolution from local convergence.