面向大规模数据的SVDD在线学习算法

针对支持向量数据描述(Support Vector Data Description,SVDD)在线学习时的支持向量数量随样本规模的扩大呈线性增加,进而导致模型更新时间呈非线性增长的问题,提出一种基于支持向量约减的支持向量数据描述(R-SVDD)在线学习方法。该算法通过执行支持向量约减,控制在线学习时的支持向量数量,从而使其具有比其他SVDD算法更快速且更稳定的模型更新时间,适合大规模数据的分类处理。首先阐述了支持向量约减的原理;进而给出了在线R-SVDD算法。在单分类和多分类数据集上的实验结果表明,R-SVDD算法相较于SVDD算法,能够在保持分类精度的基础上拥有更快的学习速度。

SVDD Online Learning Algorithm for Large-Scale Data

In order to solve the problem that the number of support vectors in support learning of support vector data description(SVDD) increases linearly with the enlargement of sample size,which leads to the nonlinear growth of model update time,an online learning method of support vector data description based on support vector reduction (R-SVDD) is proposed.By performing support vector reduction,the number of support vectors in online learning is controlled,so it has a faster and more stable model update time than other SVDD algorithms,which is suitable for large-scale data classification processing.Firstly,the principle of support vector reduction is described,and then an online R-SVDD algorithm is given.The experimental results on single and multi classification data sets show that the R-SVDD algorithm can improve the learning speed significantly while maintaining the accuracy of online SVDD classification.