摘要:本文简述了BP神经网络模型及BP学习算法规则,并在Matlab中利用BP神经网络对英文字母进行了识别,经结果分析发现BP神经网络对字母具有一定的辨识能力,是一种对字母识别的有效方法。
关键词: BP神经网络 Matlab 字母识别
Abstract:This paper first describes the back propagation network model and the rule of back propagation learning algorithm. And then identified the alphabet in Matlab using the back propagation network. The results show that the back propagation network has the ability to identify some of the letters of the alphabet, which is an effective method for character recognition.
Key Words: Back propagation network; Matlab; character recognition
1 引言
20世纪80年代中期,学者Rumelhart、McClelland和他们的同事提出了多层前馈网络MFNN(Mutltilayer Feedforward Neural Networks)的反向传播学习算法,简称BP网络(Back Propagation Network)学习算法。BP网络是对非线性可微分函数进行权值训练的多层前向网络。在人工神经网络的实际应用中,80%~90%的模型都采用BP网络或其变化形式。采用BP 算法的多层前馈网络, 是迄今为止应用最广泛的神经网络, 具有很强的非线性逼近能