The Research On Semantic Repetition Of MEP And GEP

Genetic programming (GP) becomes an important branch of evolutionary computation because of its superior ability to solve complex problems. Thus, it has been favored by many scholars,and applied to multiple areas such as artificial intelligence, financial estimation. However, GP and its branches have semantic repetition in the evolution process that seriously affect the efficiency and performance. For examples, in Multiple Expression Programming, any genes in each chromosome are likely to have been applied by genes of present or other subsequent populations, which cause the repeated calculation and waste of a lot of time and space resources. Likewise, in Gene Expression Programming, any genes in each chromosome are likely to have been applied by present genes or other effective gene segments, which influences the evolution efficiency and performance. In order to promote the evolution of MEP and GEP efficiency and performance, this paper mainly studies how to remove the semantic repeat in MEP and GEP, and to make some improvement work. Specific as follows:1)Theoretically to prove that the MEP exist the phenomenon that genes have been repeatedly used and semantic action is executed repeatedly in the process of evolution, and put forward NMEP, a new algorithm which is aimed at MEP. Under the condition of no change in MEP chromosome said rules and evolution way, NMEP can recognize the evolution process of repeated use of genes, to avoid the repeated successes and semantic actions executed repeatedly. In this way, it saves the space resources, and improves the performance of the MEP.2)Depth-preferred decoding method is applied to the GEP, and used to analyze the performance between the depth-preferred-based decoding way of GEP and the standard GEP which is based on breadth-preferred decoding way.3)Theoretically to prove that there is a phenomenon that GEP semantic actions are repeated in the evolutionary process. Thus, putting forward the new algorithm NGEP, based on the GEP which has depth-preferred decoding way. In the same time, Algorithm NGEP doesn’t change the Chromosome Expression rules and evolution way of GEP which is based on depth-preferred decoding, and is able to identify effective gene segments which are repeatedly used in the evolution process.4)By comparing the experiment compared MEP that is integrated into algorithm NGEP with the standard MEP, both two kinds of algorithms in the three experiments adopte the same mining functions and test standards. The experimental results show that algorithm NMEP is feasible, and compared with standard MEP, the MEP that is integrated with algorithm NGEP save more time.5)By comparing the experiment MEP that is integrated with algorithm NGEP with the GEP that is based on the depth-preferred decoding way and standard MEP, all of these three kinds of algorithms in the four experiments adopte the same mining functions and experimental standards. The experimental results show that the MEP which is integrated with algorithm NGEP in performance is better than the other two algorithms.