摘要
BP神经网络算法存在收敛速度慢和网络泛化能力差的缺点,影响分类识别率。为了提高网络的分类识别能力和泛化能力,在此介绍一种基于遗传算法的神经网络集成方法,即训练出多个个体BP神经网络,利用遗传算法选择差异度较大的个体BP网络进行神经网络集成,再利用该神经网络集成进行分类识别。实验结果表明,神经网络集成可以提高识别率。
Since the constringency of the BP neural network algorithm is too slow and generalization capability of neural network is not ideal, the disadvantages effect the classification identification. A method of the neural network ensemble based on the genetic algorithm is introduced for improving the classification accuracy and generalization of neural network, the way which trains several individual BP neural networks, selects those who have great variance each other to perform the neural net- work ensemble by means :of the genetic algorithm, and then carries out the classification identification with the neural network ensemble. The experimental result shows that the method can improve the identification rate.
出处
《现代电子技术》
2010年第8期148-150,共3页
Modern Electronics Technique
关键词
BP神经网络
遗传算法
神经网络集成
人耳识别
BP neural network
genetic algorithm
neural network ensemble
ear recognition