利用自适应近邻选择和低秩表示的半监督鉴别分析

Semi-supervised Discriminant Analysis Using Adaptive Neighbor Selection and Low-rank Representation

针对基于图嵌入的鉴别投影方法对近邻参数的敏感以及实际应用中样本类别信息不足对图嵌入方法鉴别性能的影响,提出一种基于自适应近邻选择和低秩表示的半监督鉴别分析方法.该方法利用所有类内样本点构造类内图来描述类内样本的紧致性,借助最远类内样本的邻域自适应地选取该邻域内不同类样本点构造类间图,以描述类间样本的可分性;此外,利用低秩表示方法挖掘不带类别信息样本的潜在低秩结构,以保留样本的全局相似关系.在ORL和FERET人脸数据库上的实验结果,验证了文中方法的有效性及对噪声的鲁棒性.

Considering the discriminant projection methods based on graph embedding are sensitive to theneighbor parameter and the fact that there is no sufficient class-label information of samples in practical applicationswhich has an impact on the performance of graph embedding based methods, a semi-supervised discriminantanalysis method based on adaptive neighbor selection and low-rank representation is proposed. The methoduses all the intraclass samples to construct the intraclass graph which can characterize the intraclass compactness,and simultaneously adaptively selects the interclass samples within the neighborhood produced by the farthest intraclasssample to construct the interclass graph which is used to characterize the interclass separability. Furthermore,the low-rank representation approach is applied to mine the latent low-rank structure of unlabeled samplesand thus preserve the global similarity relationship of samples. Experimental results on ORL and FERET face databasesdemonstrate the effectiveness of our method and the robustness to noise.