近邻渐进直推式支持向量机算法

Near Neighbor Progressive Transductive Support Vector Machine Algorithm

针对渐进直推式支持向量机算法训练速度慢和学习性能不稳定的问题,提出一种近邻渐进直推式支持向量机算法。该算法利用支持向量机中支持向量信息,选择支持向量附近的无标签样本点进行标注,采用支持向量预选取的方法减少训练集的规模,提高算法的速度。实验结果表明了该算法的有效性。

Progressive Transductive Support Vector Machine（PTSVM） has some drawbacks such as slower training speed and unstable learning performance. This paper proposes a Near Neighbor Progressive Transductive Support Vector Machine（N2pTSVM） learning algorithm. Making full use of support vectors in SVM, the method selects new unlabeled samples near support vectors. In addition, the method introduces pre-extracting support vector algorithm to reduce the calculation complexity. Experimental results show its validity.