Research And Applications Of Gene Expression Programming

Gene Expression Programming(GEP) is a new evolutionary computing model, proposed by F.Candidain 2001, which is a new self-adaptive evolutionary algorithms based on genotype and phenotype, and successfully applied to a function finding, data classification, time series, machine learning, optimal combination, image processing and other issues. GEP combines the advantages of genetic algorithm (GA) and genetic programming (GP), but also to overcome the respective shortcomings of both. In GEP the individuals are encoded as linear strings of fixed length (genotype) which are like the chromosomes of GA, and afterwards expressed as nonlinear entities of different sizes and shapes (phenotype) which are like expression trees of GP.So it can use a simple code to solve complex problems, and can easily carry out selection, crossover and mutation genetic operations. In solving many complex problems, GEP is much faster than genetic programming in 2-4 magnitudes.However, due to GEP is the only developed in recent years, GEP is a new branch of evolutionary algorithm, not as GA, GP and other algorithm has a solid mathematical foundation and A more systematic theoretical method.While creating new concepts and algorithms, F. Candida left some theoretic issues as open problems. F. Candida also uses many properties without proof. So GEP need to be more solid theoretical foundation to support algorithm .such as improving the algorithms, to improve the time efficiency and performance of GEP.By researching theory and defects of GEP, the paper main propose two kinds of improved algorithms, and experiments show that improved algorithms have a good performance.The main contents are as follows:(1) Systematically introduce the main branchs of evolutionary algorithms, including the main principles, processes, and deficiencies of GA, GP and GEP;(2) Proposed two kinds of improved version of the GEP algorithm, Clonal Selection-based Gene Expression Programming(CSGEP)and Adaptive Hierarchical Gene Expression Programming(AHGEP); (3) Experiments show that Compared with conventional GEP, the new algorithm improve the characteristics of premature convergence, time efficiency and the search performance. Besides the algorithm's success-rate has greatly improved.