快速分类神经网络在字符识别中的应用

Fast Classification Neural Net Applied to Character Recognition

本文以海明神经网络与自适应谐振理论(ART)模型学习算法为基础,从理论上分析了海明网络学习算法的缺陷,利用ART网络的思想,提出了一种快速分类的神经元网络的算法,命名为Improved Hamming算法(简称Im-H算法)。此算法主要优点在于阈值更新及引入了经验迭代次数。将此算法用于字符模式识别,大量的计算机实验结果表明了Im-H网络学习算法的有效性、快速性。

This paper is based on Hamming memory neural network and Adaptive Resonance Theory (ART) model. In theory, we analyzes defects of the Hamming network algorithm. We utilized the idea of ART and developed a fast classification algorithm, called improved hamming algorithm (Im-H Algorithm). This algorithm provides two important properties of updating the thresholds and imtroducing empirical iterations. We have applied this methed to character recognition. A large number of experiments demonstrate its efficiency and its high covergence rate.