A Parallel Approach for Entropy-based Micro GA

In this paper, the advantage of entropy is analyzed firstly based on the prior information entropy-based genetic algorithm. then a micro-GA is presented and subsequently introduced its parallel implementation with coarse grain. The so called micro-GA is a GA with micro-population scheme. Taking advantage of the merit of multi-population, population size can be cut down appropriately by means of inter-population crossover. Because of the inter-population operator, the individualspsila diversity will not turn worse due to the shrunken population size. The parallel strategy comprises a mapping of one (or a few) population(s) onto each processor of MIMD multiprocessing system. Both the micro and parallel approach can speed up the whole genetic evolutionary procedure. Numerical examples and the performance test show that the proposed method has good accuracy and efficiency.