Some Tucker-like approximations based on the modal semi-tensor product

Approximating higher-order tensors by the Tucker format has been applied in many fields such as psychometrics, chemometrics, signal processing, pattern classification, and so on. In this paper, we propose some new Tucker-like approximations based on the modal semi-tensor product (STP), especially, a new singular value decomposition (SVD) and a new higher-order SVD (HOSVD) are derived. Algorithms for computing new decompositions are provided. We also give some numerical examples to illustrate our theoretical results.