有样本缺失的稀疏交叉视图的多重集典型相关分析

从融合多组特征的角度出发,以多重集典型相关分析算法(MCCA)为研究基础,通过稀疏保持自适应选择样本局部信息,然后通过在同类样本之间计算权重矩阵,将样本类别信息嵌入到算法中,再利用多种视图之间的交叉相关项,克服不同视图样本必须成对出现的局限,提出一种有样本缺失的稀疏交叉视图的多重集典型相关分析算法(multiset canonical sparse cross-view correlation analysis with missing samples,CSMCCAM)。在手写体数据集和CENPARMI数据库上验证本文的算法,得到CSMCCAM算法分类精确度优于LPMCCAM等典型相关分析算法,并且对缺失样本数目不敏感。

Multiset canonical sparse cross-view correlation analysis with Missing Samples

Based on multiset canonical correlation analysis in the fusion of multi group feature, using sparsity preserving to select local information adaptively, through calculating the weight matrix in samples of the same class embedded the class information into the algorithm, at last introduce the cross-view correlation to overcome the limitations of different view samples must come in pairs, then the multiset canonical sparse cross-view correlation analysis with Missing Samples (CSMCCAM) is proposed. The experimental results on the multiple feature database and CENPARMI database show that the proposed CSMCCAM outperforms the related canonical correlation analysis recognition methods and the recognition accuracy of CSMCCAM is relatively insensitive to the number of missing samples.