Eigenvector-based sparse canonical correlation analysis: Fast computation for estimation of multiple canonical vectors

Classical canonical correlation analysis (CCA) requires matrices to be low dimensional, i.e. the number of features cannot exceed sample size. Recent developments in CCA have mainly focused on high-dimensional setting, where both under greatly exceeds These approaches impose penalties optimization problems that are needed solve iteratively, and estimate multiple vectors sequentially. In this work, we provide an explicit link between sparse regression with analysis, efficient algorithm can pairs simultaneously rather than Furthermore, naturally allows parallel computing. properties make much efficient. We theoretical results consistency pairs. The development based solving eigenvectors problem, which significantly differentiate our method existing methods. Simulation support improved performance proposed approach. apply eigenvector-based GTEx thyroid histology images, SNPs RNA-seq gene expression data, a microbiome study. real data also shows compared traditional CCA.