Sparsifying the least-squares approach to PCA: comparison of lasso and cardinality constraint

Abstract Sparse PCA methods are used to overcome the difficulty of interpreting solution obtained from PCA. However, constraining obtain sparse solutions is an intractable problem, especially in a high-dimensional setting. Penalized due their computational tractability. Nevertheless, recent developments permit efficiently obtaining good cardinality-constrained problems allowing comparison between these approaches. Here, we conduct penalized method with its counterpart for least-squares formulation imposing sparseness on component weights. We compare and through simulation study that estimates structure’s recovery, mean absolute bias, variance, squared error. Additionally, use data set illustrate practice. Results suggest using leads better recovery structure.