基于遗传算法的神经网络权值优化  

Optimization of Weights in Neural Networks Based on Genetic Algorithm

作　　者：洪露[1] 马长山[1] 谢宗安[1] 

机构地区：[1]贵州工业大学电气工程学院,贵州贵阳550003

出　　处：《贵州工业大学学报（自然科学版）》2003年第6期48-51,共4页Journal of Guizhou University of Technology(Natural Science Edition)

摘　　要：针对BP算法学习效率低、收敛速度慢,以及易陷入局部最优等缺点,提出了一种新型的、基于自然选择和基因遗传学原理的随机搜索算法——遗传算法(Genetic Algorithm),并论述了它在BP神经网络中权值优化的问题。仿真结果表明,用遗传算法优化BP神经网络的权值收敛速度快,并有效解决了BP算法易陷入局部最优的问题。Aimed at the shortcomings of the BP algorithm,such as low efficiency of study,slow speed of convergence,easy to be trapped into local optimums,etc,this paper introduces Genetic Algorithm (GA) -a new algorithm based on natural selection and the principle of gene genetics, and discusses its weights optimizing in BP neural networks. The simulation results show that GA can improve the convergence speed effectively and avoid the trap of local optimums.

关 键 词：遗传算法 BP神经网络 权值 BP算法 GA 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]