机构地区: 中南大学信息科学与工程学院
出 处: 《计算机工程与应用》 2003年第19期29-30,50,共3页
摘 要: 该文针对BP神经网络易限入局部极小的问题,提出了混沌梯度优化的神经网络的学习算法,其原理是:用规则来判断由于梯度搜索过程中产生的局部极小,并利用具有全局寻优的特点的混沌搜索,使学习过程能有效地逃离局部极小。即采用梯度下降进行“粗搜索”,混沌搜索进行“细搜索”,并建立规则将两者结合起来,就构成了BP神经网络的基于规则的混沌梯度耦合学习算法。它有效地利用了梯度下降算法的快速性和混沌寻优的全局性,并已应用于工程实际,取得了良好的效果。 Aiming at the problem that BP algorithm is usually to produce local minimum,in this paper,some rules to judge local minimum in BP algorithm is proposed,and an improved chaotic optimization algorithm is proposed to jump local minimum effectively.This coupled algorithm is made up of raw searching by gradient ,elaborate searching by chaotic searching and some rules to judge local minimum.It makes full use of quickness of gradient search and full scope search of chaotic optimization.Its practical application shows that the algorithm can resolve the problem on exercise efficiency of BP algorithm.
领 域: [自动化与计算机技术] [自动化与计算机技术]