基于遗传算法演化网络的字符识别研究  

Research of Character Recognition Based on Neural Networks Evolved by Genetic Algorithms

作　　者：林万洪[1] 薛亮[1] 

机构地区：[1]航天医学工程研究所,北京100094

出　　处：《电子科技》2005年第5期13-16,共4页Electronic Science and Technology

基　　金：总装试验计划项目资助

摘　　要：将神经网络应用于字符识别,对BP网络和经遗传算法演化后的BP网络进行比较研究.计算结果表明,在训练效果与识别效果两个方面,遗传算法演化后的网络具有更好的识别效果,并能够大幅度提高网络训练时的收敛速度.通过理论分析其原因后得出结论,利用遗传算法做字符识别具有更优的性能.This paper applies artificial neural network to character recognition and compares the training and recognizing effect between the simple BP algorithms and the BP networks evolved by genetic algorithms. The result shows that networks evolved by genetic algorithms have better recognition ability, moreover, it can improve the convergence rate of network training. Based on the theory analysis, we come to the conclusion that character recognition system with genetic algorithms has better performance.

关 键 词：遗传算法 神经网络 字符识别 BP算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]