The Strategies to Improve Performance of Function Mining By Gene Expression Programming -Genetic Modifying, Overlapped Gene, Backtracking and Adaptive Mutation

This paper introduces the technologies to improve the performance of function mining by Gene Expression Programming (GEP) developed in Sichuan University last year. The main results include: (a) Genetic Modifying Algorithm (Trans-gene). By injection gene segment into genome, it guides the evolutional direction and speeds up knowledge discovery process. (b) Overlapped gene expression. Borrowing the idea of overlap gene expression from biological study, it applies overlapped gene expression, saves space for gene expression. (c) Backtrack-able GEP. Enlightened by atavism in biology, it proposes backtrack-able GEP algorithms, designing Geometric Proportion Increased Checkpoint Sequence and Accelerated Increased Checkpoint Sequence to restrict the backtrack process. (d) Adaptive Mutation. The mutation rate for each individual can vary in evolution according to the value of fitness. Experiments show that these techniques boost the performance of GEP by one or two magnitudes, respectively. Keyword Template Knowledge discover, Gene Expression Programming, Function Mining 1 This work was supported by the National Science Foundation of China under Grant No.60473071, the National Research Foundation for the Doctoral Program by the Chinese Ministry of Education under Grant No.20020610007.