On the Largest Singular Value/Eigenvalue of a Random Tensor

Eigenvalues, singular values, and eigenvectors (15A18) Inequalities involving eigenvalues and eigenvectors (15A42) Multilinear algebra, tensor calculus (15A69) Gaussian processes (60G15)

arXiv classification

math.OC math.PR math.SP

Abstract: This short note presents upper bounds of the expectations of the largest singular values/eigenvalues of various types of random tensors in the non-asymptotic sense. For a standard Gaussian tensor of size

n_{1} i m e s c d o t s i m e s n_{d}

, it is shown that the expectation of its largest singular value is upper bounded by

s q r t n_{1} + c d o t s + s q r t n_{d}

. For the expectation of the largest

e l l^{d}

-singular value, it is upper bounded by

2^{f r a c d - 1 2} p r o d_{j = 1}^{d} n_{j}^{f r a c d - 2 2 d} s u m_{j = 1}^{d} n_{j}^{f r a c 12}

. We also derive the upper bounds of the expectations of the largest Z-/H-(

e l l^{d}

)/M-/C-eigenvalues of symmetric, partially symmetric, and piezoelectric-type Gaussian tensors, which are respectively upper bounded by

d s q r t n

,

d c d o t 2^{f r a c d - 1 2} n^{f r a c d - 1 2}

,

2 s q r t m + 2 s q r t n

, and

3 s q r t n

.

This page was built for publication: On the Largest Singular Value/Eigenvalue of a Random Tensor

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6370239)