扩散遗传算法及其应用研究

A Diffusing Operator Based Genetic Algorithm

针对遗传算法在局部搜索能力方面的缺陷,提出了一种基于扩散算子的遗产算法(简称扩散遗产算法)。该算法中包含的扩散算子是变异算子,其主要作用是在遗传搜索中进行局部搜索。用扩散遗传算法和实数编码遗传算法分别训练用于解XOR问题的神经网络,对比结果表明,论文提出的算法兼具强的全局搜索能力和局部搜索能力,因此,该算法可以不借助其它局部搜索算法而单独作为神经网络训练算法,从而简化训练算法,提高训练效率。该算法对提高遗传算法搜索效率和求解精度具有重要的意义。

A diffusing operator based genetic algorithm（DOBGA） is proposed,in which a diffusing operator is designed, in order to improve the local search ability of a genetic algorithm.Gaussian mutation method is employed in diffusing operator,which mainly performs short-step local search in genetic algorithm.Connection weights of an artificial neural network are trained on standard XOR problem by using the proposed genetic algorithm.The results show that the proposed genetic algorithm can perform both global search and local search perfectly,therefore,it can be used to train artificial neural networks alone rather than incorporate other local search algorithms,such as BP to improve local search ability of a training algorithm,so the diffusing operator based genetic algorithm is significant to simplify the training algorithm of artificial neural networks and improves training efficiency greatly.