Improvement of Interprocessor Communication in A Parallel Genetic Algorithm Retaining Sequential Behaviours

This paper proposes an improvement of interprocessor communication in a parallel genetic algorithm maintaining the execution behaviours of sequential genetic algorithms. The global population is evenly partitioned into several subpopulations each of which is assigned to the processor. An interprocessor communication at each generation in order to exchange the information on all individuals is improved by two steps: First only information on fitness value is exchanged between processors. According to these fitness values, the individuals for crossover can be selected. Then the chromosomes of these selected individuals are exchanged between processors, if necessary. Some experiments on AP1000 show the efficiency of this improvement. 1 Introduction Genetic algorithms are devised by imitating the natural selection, genetic crossover and mutation found in the evolution process of the living systems [1]. Genetic algorithms are expected as powerful solving method for various practical proble..