改进自适应遗传算法在BP神经网络学习中的应用  

Application of Improvement Self-Adaptation Genetic Algorithm in BP Neural Network Learning

作　　者：李孝忠[1] 张有伟[1] 

机构地区：[1]天津科技大学计算机科学与信息工程学院,天津300222

出　　处：《天津科技大学学报》2010年第4期64-67,共4页Journal of Tianjin University of Science & Technology

基　　金：国家自然科学基金资助项目(70571056)

摘　　要：针对遗传算法容易产生局值的问题,提出一种新的自适应遗传算法,改进遗传算子,通过比较两代之间的适应度评估值,选取适合的交叉率和变异率,保证了优秀个体进入下一代,而且避免了种群中最大适应度值的个体的交叉率和变异率为0的情况.最后,将改进后的算法应用于库存控制模型,实验表明,改进后的自适应遗传算法能避免局值,提高网络的收敛速度,改善了网络的学习性能.A new self-adaptable genetic algorithm was put forward to solve the problem that it was easily to sink into the partial minimum in the genetic algorithm.By means of improving genetic operators and comparing the fitness value between two generationst,he appropriate crossover rate and mutation rate can be selected to ensure the excellent individual into the next generation.And the situation that the largest population fitness value of individual cross-rate and mutation rate were zero can be avoided.Lastlyt,he improved algorithm was applied to the inventory control model.Results show that the improvement self-adaptation genetic algorithm can avoid sinking into partial minimume,nhance the convergence rate of the networka,nd improve the network study performance.

关 键 词：BP神经网络 遗传算法 自适应 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]