基于混沌蚁群算法的BP神经网络训练研究  

The Research of BP Neural Network Training Based on the Chaos Ant Colony Optimization

作　　者：袁冬根[1] 刘晓东[1] 蔡磊[1] 袁德国[1] 

机构地区：[1]空军工程大学工程学院,陕西西安710038

出　　处：《微电子学与计算机》2009年第4期11-14,共4页Microelectronics & Computer

摘　　要：针对BP学习算法依赖于初始权值的选择和只能处理具有求导或者梯度特性目标函数,导致函数逼近误差较大的缺点,基于适值理论,建立了混沌蚁群算法模型.采用实值编码,把BP神经网络的权值和偏值作为混沌蚁群算法搜索的蚂蚁空间位置;设置合适的适值函数,对BP神经网络权值进行训练,并将其应用于曲线拟和.分别用混沌蚁群算法对不包含噪声的BP神经网络和包含噪声的BP神经网络进行了训练仿真,对仿真结果进行分析,取得了很好的效果.The algorithm model of Chaos Ant Colony Optimization （CACO） is built up based on fitness theory. The method of value coding is adopted; the power value and partial value of the BP neural network are taken as ant space position which is searched by CAC_ID. The suitable fitness function set, this paper proposes the new seherne of using CACO to train the BP neural network and applies the CACO- designed BP neural network into the problems of function approaching and sorting. BP neural network is simulated and educated by CACO, and the results are analyzed, which is obtained better effect.

关 键 词：群智能 混沌蚁群算法 BP神经网络 

分 类 号：E927[军事—军事装备学]