Research Of Function Clustering And Evolutionary Computation Knowledge Acquisition

The research of the basic components in the genetic algorithm,such as selection, crossover,mutation operators and population size,truncated generation,crossover probability, mutation probability and so on, show that the choice of parameters can have significant impact on the effectiveness of the optimization algorithm:At present,the optimal control parameters suggestions either are too wide range or each are not identical. For solving this problems,in this dissertation, a novel clustering based on fitness landscape is proposed to classify functions,and then using the class information of functions and knowledges baesed on them to guide optimization algorithm.The studies focused on and major achievements made in this dissertation are as follows.1) propose a novel function oriented clustering representation.On the basis of function fitness landscape,a novel clustering representation is proposed and the correspond clustering algorithms is also constructed. The experimental results show that the clustering algorithm not only satisfys clustering level effectiveness,and meets the effectiveness of evolutaionary algorithms level.2) design a experimental platform can be flexible extented and construct a experimental results database.valuating large amounts of experimental results are achieved under the experimental platform,a experimental result database is constructed which are not only using to validate the clustering results,but also provides the basis for knowledge discovery based on the class information of functions.3) knowledge disco very. the experimental results show that these knowledges can improve convergence accuracy and convergence probability and reduce convergence generation.