Tubal Matrix Analysis

05/03/2021
by   Liqun Qi, et al.
0

One of the early ideas started from the 2004 workshop is to regard third order tensors as tubal matrices such as matrix theory can be naturally extended to third order tensors. We revisit this approach in this paper. We define the value of a tubal scalar to remedy the gap between this approach and the practical needs to have singular values as real scalars. We define the transpose of a tubal scalar such that the transpose of a tubal matrix is a natural extension of the transpose of a matrix and still corresponding to the transpose of a third order tensor. We define positive semi-definite symmetric tubal scalars such that s-diagonal tubal matrices can be defined meaningfully and are still corresponding to third order s-diagonal tensors, which were introduced recently for studying the T-SVD factorizations. We show that the 2-norm of a tubal matrix is equal to its largest T-singular value, multiplied with a coefficient, which is 1 in the case of matrices. This shows that T-singular values of tubal matrices are natural extensions of singular values of matrices. Further study on tubal matrices may reveal more links between matrix theory and tensor theory.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
03/01/2021

T-Singular Values and T-Sketching for Third Order Tensors

Based upon the T-SVD (tensor SVD) of third order tensors, introduced by ...
research
12/23/2021

When Random Tensors meet Random Matrices

Relying on random matrix theory (RMT), this paper studies asymmetric ord...
research
05/30/2017

A general theory of singular values with applications to signal denoising

We study the Pareto frontier for two competing norms ·_X and ·_Y on a ve...
research
07/17/2020

Perturbation Bounds for Orthogonally Decomposable Tensors and Their Applications in High Dimensional Data Analysis

We develop deterministic perturbation bounds for singular values and vec...
research
04/12/2021

ST-SVD Factorization and s-Diagonal Tensors

A third order real tensor is mapped to a special f-diagonal tensor by go...
research
08/14/2022

The SVD of Convolutional Weights: A CNN Interpretability Framework

Deep neural networks used for image classification often use convolution...
research
03/24/2021

On the ℓ^∞-norms of the Singular Vectors of Arbitrary Powers of a Difference Matrix with Applications to Sigma-Delta Quantization

Let A _max := max_i,j |A_i,j| denote the maximum magnitude of entries of...

Please sign up or login with your details

Forgot password? Click here to reset