论文部分内容阅读
模式识别是神经网络最有前景的应用领域之一,本文主要讨论如何提高多层神经网络 BP(Back-propagation)算法的学习速度以及该算法用于手写数字识别的研究.文中提出了局部连接的网络结构,并对基于特征输入和基于点阵输入两种神经网络分类器的特点进行了比较,针对神经网络的识别机制、识别能力和自适应学习,进行了深入讨论.本文还给出容错能力的概念,用以描述神经网络对非学习样本的分类机制.所有研究工作是在作者研制的 SSNN 神经网络仿真软件上进行的.
Pattern recognition is one of the most promising application fields of neural networks, this article mainly discusses how to improve the learning speed of BP neural network (BP-Back-propagation) algorithm and its application in handwritten numeral recognition.In this paper, Network structure, and compares the characteristics of two kinds of neural network classifiers based on feature input and dot matrix input. The recognition mechanism, recognition ability and adaptive learning of neural network are discussed in depth.This paper also gives fault tolerance , Which is used to describe the classification mechanism of non-learning samples by neural networks.All research work is carried out on the SSNN neural network simulation software developed by the author.