The Research On Gene Expression Programming Using Depth-First And Breadth-First Decoding Principle And Its Application

Gene Expression Programming(GEP) combined with advantages of Genetic Algorithm(GA) and Genetic Programming(GP), it can represent the expression tree with tree structure of different shape and size by using chromosome of linear fixed length. Compared with traditional GP, the genotype and phenotype of GEP is separated, it makes the genetic operation more convenient. It also can make the algorithm search the whole solution space more quickly, which makes algorithm has faster convergence.However, the GEP algorithm has proposed in 2001, the research on GEP algorithm and its application has just beginning. So the algorithms need some more theory to support. Because the genotype and phenotype of GEP is separated, there is no correspondence relationship between sub-trees in the expression trees and contiguous segments in the linear chromosome, and the solution structures are very fragile when subject to genetic operations. We propose a new representation scheme that overcomes the original GEP's drawbacks, and apply the GEP algorithm to the image retrieval. The main works as follows:1) Based on the research of the advantages and disadvantages and the core technology of traditional GEP algorithm, a new decoding method(combined the depth-first and breadth-first decoding principle of GEP, DBGEP) is proposed. Combining with two different decoding principles, it can get different expression tree(ET) with the same expression. The experimental showed that, compared with the standard GEP algorithm, the new algorithm can improve the mean fitness, leading the algorithm to a higher success rate.2) This paper research on the core technology and problem of content based image retrieval. In order to solve the problem that the retrieval accuracy of use single feature or even more features is not high, the GEP algorithm is applied to the image retrieval. Through the GEP algorithm, a suitable nonlinear combination function is generated adaptively, and the image retrieval is realized, and the accuracy of content based image retrieval is improved.