摘要
粗糙集和神经网络在模式识别中都可用于分类,但是都有局限性。虽然两者没有太多的共同点,将它们结合起来却能相互补充,起到比单个理论更好的分类效果。本文从理论上给出了用粗糙集约简算法减少BP网络中的一个神经元或连接时网络输出能产生的最大误差。接着将粗糙集和 BP网络结合起来设计分类器,并通过车牌数字识别验证了该分类器的有效性。实验说明该分类器比单独用粗糙集和神经网络设计的分类器识别率高、识别时间短。
In .pattern r egconition, both rough set theory and neural network can be used to classify patterns. The two theory, don＇t have much commonpoint, but they .can be. complementary when we combine them. in designing a classifier and the recognition effect can be better. This article gives the max output error when the rough .setis used to reduce a nerve cell or a connection in Back-Propogation（BP）neural network.Then, rough set theory and BP network are integrated into one classifier.we validate this classifier＇s .validity by using it in the character recognition of lieence plate. The exp.eriment.shows that this classifier is better than classifiers using rough set or BP neural network in recognition rate and recognition speed.
出处
《计算机科学》
CSCD
北大核心
2005年第11期172-174,共3页
Computer Science
基金
国家自然科学基金(60175016
60475019)
