摘要
为了更好地利用自适应共振理论(ART1)来实现汉字的分类,提出了一种改进的ART1算法。改进算法引用了同或的思想,将输入模式与记忆模式相同的部分在输入模式总体中占的比例作为二者的匹配度,从而降低了输入样本顺序对分类结果造成的误差,增强了网络的稳定性和抗噪性。实验结果证明,改进算法能够更好地将汉字进行分类,求出聚类中心。
In order to make better use of adaptive resonance theory （ART1） to achieve the classification of Chinese characters, an improved algorithm of ART1 was proposed. The algorithm quotes the thought of XNOR, the percentage of the same part between the input and the memory mode in the overall input mode was used to define the match degree, thus the classification errors of the input sample sequence is reduced and the stability and the noise immunity of the network are enhanced. The experimental results proved that the improved algorithm can classify Chinese characters better and the cluster center can be calculated.
出处
《北京石油化工学院学报》
2009年第1期38-42,共5页
Journal of Beijing Institute of Petrochemical Technology
关键词
神经网络
自适应共振理论
警戒参数
聚类中心
artificial neural network
adaptive resonance theory
vigilance parameter
clustercenter
