面向大规模数据的隐私保护学习机

随着海量数据不断涌入,SVM隐私泄露问题日益严重。在分析已有隐私保护支持向量机基础上,提出一种面向大规模数据的隐私保护学习机(PPLM)。该方法首先通过核心向量机对大规模样本进行采样,然后在核心集上选取两个样本点并将两点连线的法平面作为最优分类面。通过对标准数据集和人工数据集的实验表明,PPLM可有效地解决大规模样本分类问题,且分类效果良好。

Privacy Preserving Learning Machine for Large Scale Datasets

Support vector machine (SVM) is widely used in pattern classification. In order to solve the privacy preserving problem in SVM, a privacy preserving learning machine for large scale datasets (PPLM) is proposed in this paper. First, core vector machine (CVM) is introduced for sampling the large scale datasets; then two points from different classes are chosen in the core set and the hyperplane orthogonal to the line connecting these two points is treated as the optimal separating hyperplane. Experimental results obtained from synthetic and standard datasets verify that the PPLM is effective and competitive.