首页> 中文期刊>中国科技论文 >基于 AP 和 BP 神经网络算法的手写数字识别

基于 AP 和 BP 神经网络算法的手写数字识别

摘要

Given the problem that current methods of handwritten digit recognition are not ideal for large-scale application,a new method of handwritten digit recognition has been proposed,combining affinity propagation with error back-propagation neural network algorithm.Firstly,pretreatment of samples was carried out.Then the AP algorithm was used to cluster samples to e-liminate redundant and re-construct the sample space.Finally,the BP neural network was utilized to learn and recognize each class from AP clustering.Experiments were conducted with the data from UCI machine learning database,and the correct identi-fication rate of the method reaches 96.10%,which is better than that of BP neural network algorithm (94.88%),and the pro-cessing time of the method is only one over ten of BP neural network algorithm.Thus,the proposed method can be used to effi-ciently and effectively identify handwritten digits with high practical value.%针对现有的手写数字识别技术不适合大规模应用的问题,提出了一种基于 AP 和 BP 神经网络的快速手写数字识别算法。首先对预处理后的样本通过 AP 算法(affinity propagation)聚类消除冗余,重新构造样本空间;然后构造 BP(误差反向传播)神经网络模型,学习测试集合样本。采用 UCI 机器学习数据库中的数据进行实验,结果表明,算法的识别正确率可达96.10%,高于 BP 神经网络算法的识别正确率94.88%,且执行时间约为后者的10%,具有较高的实用价值。

著录项

相似文献

  • 中文文献
  • 外文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号