基于Tri-training直推式支持向量机算法

Transductive Support Vector Machine Based on Tri-training

针对直推式支持向量机错误累积及获取无标记样本空间信息慢的问题,结合Tri-training算法、KKT条件及富信息策略提出一种基于Tri-training的直推式支持向量机算法,用KKT条件选择标注样本,用富信息策略选择加入的分类器,利用多个分类器的投票结果进行标注,提高样本标注的准确度,利用多个分类器进行协同训练提高算法的训练速度.最后实验结果表明,算法能够提高最终分类器的分类精度和算法的训练速度.

Abstract： For the problem of the error accumulation and slowly obtaining the space information from unlabeled samples,combined with Tri- training algorithm,KKT condition and rich information strategy,transductive support vector machine algorithm is proposed based on Tri-training. This algorithm selects the labeled samples with KKT condition,selects the classifiers with rich information strategy and marks the unlabeled samples with voting results. This algorithm can improve the training speed and classification accuracy. Finally,the experimental results show that the proposed algorithm can improve the classification accuracy and the training speed.