Research on Fuzzy Adaptive Search Method for Parallel Genetic Algorithms

It is a problem to GA that the search ability of the ordinary GA is not always optimal specially in the early and final of the search stage, because genetic parameters (crossover rate, mutation rate and so on) are fixed. To order overcome this problem, our labo-ratory have already proposed Fuzzy Adaptive Search method for Genetic Algorithm (FASGA) as the mod-ified method [1]. This method is able to realize the efficient search by describing of fuzzy rules to tune genetic parameters according to the search stage. On the other hand, some GAs in parallel methods were already proposed as the effective method for finding high quality solution [2]. The disadvantage of PGA is that it is not always effective by using parallel process-ing because there are many parameters as compared with ordinary GA and the migration of individuals are not necessarily performed only in case of neces-sity. Therefore, in this research, we proposed Fuzzy Adaptive Search method for Parallel Genetic Algo-rithm (FASPGA) combined FASGA with PGA[3]. It is the method with the ability search high-quality so-lutions in fast which is tuned the crossover, mutation and migration rate by fuzzy reasoning. In the end, we simulated with Rastrigin function by computer for comfirming the performance of FASPGA proposed in this paper. 2 FASPGA 2.1 Parameters As above, we know that the whole genetic param-eters of FASPGA include in crossover rate rci, mu-tation rate rmi and migration degree rei are decided by fuzzy rule. The membership function in IF part in FASPGA we proposed is same to FASGA. It is composed of the average fitness value fai and the difference between the maximum and average fitness value (fmi-fai). But, FASPGA add a new paramete