摘要
为了将一般增量学习算法扩展到并行计算环境中,提出一种基于多支持向量机分类器的增量学习算法.该算法根据多分类器对新增样本集的分类结果,以样本到分类超平面的平均距离为条件重新构造支持向量集更新分类器,直到所有分类器的分类精度满足指定阈值.实验结果表明了该算法的可行性和正确性.
In order to extend common incremental learning algorithms into a parallel computation setting, an incremental learning algorithm with multiple support vector machine classifiers is proposed. According to the results of multiple classifiers, new samples were selected to be support vectors sets by computing the distance mean of the samples to the hyperplane, until all classifiers were updated and all classification accuracies met the given threshold. The experiment results on test data sets prove the feasibility and validity of the proposed algorithm.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2006年第1期103-106,共4页
Journal of Harbin Engineering University
基金
黑龙江省自然科学基金资助项目(F2005-02)
