$p$-order Tensor Products with Invertible Linear Transforms

TL;DR

This paper introduces a generalized higher order tensor product with invertible transforms, explores its properties, and connects it to tensor rank and nuclear norm, facilitating future low-rank tensor recovery methods.

Contribution

It proposes a new generalized tensor product with invertible transforms and establishes its properties and relation to tensor rank and nuclear norm.

Findings

Introduces a generalized tensor product with invertible transforms.

Proves properties of the new tensor product.

Links tensor nuclear norm to multi-rank, aiding low-rank recovery.

Abstract

This paper studies the issues about tensors. Three typical kinds of tensor decomposition are mentioned. Among these decompositions, the t-SVD is proposed in this decade. Different definitions of rank derive from tensor decompositions. Based on the research about higher order tensor t-product and tensor products with invertible transform, this paper introduces a product performing higher order tensor products with invertible transform, which is the most generalized case so far. Also, a few properties are proven. Because the optimization model of low-rank recovery often uses the nuclear norm, the paper tries to generalize the nuclear norm and proves its relation to multi-rank of tensors. The theorem paves the way for low-rank recovery of higher order tensors in the future.