江汉大学学报(自然科学版) ›› 2016, Vol. 44 ›› Issue (5): 469-473.doi: 10.16389/j.cnki.cn42-1737/n.2016.05.015

• 计算机科学 • 上一篇    下一篇

支持向量机在线增量学习算法的MATLAB 实现

杨志红,隆涛   

  1. 江汉大学 物理与信息工程学院,湖北 武汉 430056
  • 出版日期:2016-10-28 发布日期:2016-11-04
  • 作者简介:杨志红(1977—),男,副教授,博士,研究方向:模式识别、智能控制。

Realization of Online Incremental Learning Algorithm of Support Vector Machine Based on MATLAB

YANG Zhihong,LONG Tao   

  1. School of Physics and Information Engineering,Jianghan University,Wuhan 430056,Hubei,China
  • Online:2016-10-28 Published:2016-11-04

摘要: 基于支持向量机在线训练理论,采用MATLAB 工具实现了一种在线增量学习算法。该实现方法调用LIBSVM 工具包分类器训练函数和样本识别函数,将不能识别的样本作为增量数据重新训练分类器。设计的增量训练用户图形界面系统应用于手写数字的识别,结果表明了实现方法的正确性和有效性,并论证了增量训练能在兼顾新样本出现的情况下提高训练速度,提升手写数字识别的准确率。

关键词: 支持向量机, 在线增量学习算法, MATLAB, 手写数字识别

Abstract: An online incremental learning algorithm is implemented with MATLAB based on the online training theory of support vector machine. This method uses the unidentified samples as incremental data to train classifier through svmtrain and svmpredict functions of LIBSVM toolkit. The results of handwritten numeral recognition GUI system demonstrate the correctness and effectiveness of the design,moreover,the incremental training can improve the training speed and improve the accuracy of handwritten numeral recognition in the case of taking into account the emergence of new samples.

Key words: support vector machine, online incremental learning algorithm, MATLAB, handwritten numeral recognition

中图分类号: