基于重复训练提高SVM识别率的算法

持向量机 (supportvectormachine ,SVM)作为一种新的模式识别算法 ,在许多模式识别问题上表现出了良好的识别性能和推广能力。和其它模式识别方法一样 ,如何进一步提高识别率一直是研究的热点。通过分析SVM的分类机理 ,提出了一种基于重复训练的SVM算法。该算法以较小的训练代价 ,通过提高支持向量在样本中的比例 ,提高了SVM的识别性能。基于重复训练的SVM算法为提高识别率提供了一种新的思路 ,也为SVM算法的自学习奠定了基础。

An Algorithm to Improve the Recognizing Rate Based on Retraining Samples

As a new algorithm of pattern recognition, support vector machine (SVM) has a good performance in recognizing and generalizing. Like other algorithms, the method to improve its performance is a hotspot. A new algorithm based on retraining samples is presented, performance of recognition will improve at a little expense by increasing the proportion of support vectors in training samples. The algorithm based on the retraining samples provides a new idea for improving recognizing rate, and lays the foundation for auto-learning of SVM.