支持向量机在线增量学习算法的MATLAB实现

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

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。

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.