Genetic Algorithm, A Number Of Improvements And Applications

Genetic Algorithm is a searching method using probability and its basic theorem is to simulate the biology evolution.GA do not require the solution space to be continuious anddiravable.Its few limitation on the presumption of the solution space and built-in parallelism makeit a useful method in many area. This article briefly summarizes the rationales of GeneticAlgorithm and gives an intensive discussion about the premature convergency, genetic driftingand society diversity. Based on the discussion, I designed Genetic Algorithm to solve the problemof function optimization. The algorithm utilize the "share function" to measure the societydiversity which in turn decide frequency of crossover operation and mutation operation.In order tosolve the TSP, we present a new crossover operator by using the theory of "beneficial crossover ".With this new operator, we build up a heuristic searching strategy. Experiments show that there aremany valuable "crumbs" in the solution which can be combined into good solution by exchangingthe position of these "crumbs". So we design a "crumb switching" mutation operator which caneffectively improve the possibility of the right combination of these "crumbs".