Abstract:Various iterative eigenvalue solvers have been developed to compute parts of the spectrum for a large sparse matrix, including the power method, Krylov subspace methods, contour integral methods, and preconditioned solvers such as the so called LOBPCG method. All of these solvers rely on random matrices to determine, e.g., starting vectors that have, with high probability, a non-negligible overlap with the eigenvectors of interest. For this purpose, a safe and common choice are unstructured Gaussian random matrices. In this work, we investigate the use of random Khatri-Rao products in eigenvalue solvers. On the one hand, we establish a novel subspace embedding property that provides theoretical justification for the use of such structured random matrices. On the other hand, we highlight the potential algorithmic benefits when solving eigenvalue problems with Kronecker product structure, as they arise frequently from the discretization of eigenvalue problems for differential operators on tensor product domains. In particular, we consider the use of random Khatri-Rao products within a contour integral method and LOBPCG. Numerical experiments indicate that the gains for the contour integral method strongly depend on the ability to efficiently and accurately solve (shifted) matrix equations with low-rank right-hand side. The flexibility of LOBPCG to directly employ preconditioners makes it easier to benefit from Khatri-Rao product structure, at the expense of having less theoretical justification.

Solving a Class of Infinite-Dimensional Tensor Eigenvalue Problems by Translational Invariant Tensor Ring Approximations

A Flexible Power Method for Solving Infinite Dimensional Tensor Eigenvalue Problems

On low-rank approximability of solutions to high-dimensional operator equations and eigenvalue problems

An Inexact Cayley Transform Method for Inverse Eigenvalue Problems with Multiple Eigenvalues

An Ulm-like Cayley Transform Method for Inverse Eigenvalue Problems with Multiple Eigenvalues

Approximate Cayley Transform Methods for Inverse Eigenvalue Problems and Convergence Analysis

Preconditioned low-rank Riemannian optimization for linear systems with tensor product structure

Subspace Methods with Local Refinements for Eigenvalue Computation Using Low-Rank Tensor-Train Format.

A tensor approximation method based on ideal minimal residual formulations for the solution of high-dimensional problems

The infinite Lanczos method for symmetric nonlinear eigenvalue problems

Inflationary dynamics for matrix eigenvalue problems

The Waveguide Eigenvalue Problem and the Tensor Infinite Arnoldi Method

Low-rank approximate inverse for preconditioning tensor-structured linear systems

The Convergence of Harmonic Ritz Values, Harmonic Ritz Vectors, and Refined Harmonic Ritz Vectors

Approximating matrix eigenvalues by subspace iteration with repeated random sparsification

On The Tensor Svd And The Optimal Low Rank Orthogonal Approximation Of Tensors

Exact Eigenvalues and Eigenvectors for Some n-Dimensional Matrices

An Analysis of the Rayleigh-Ritz Method for Approximating Eigenspaces

On the Convergence of Ritz Pairs and Refined Ritz Vectors for Quadratic Eigenvalue Problems

Subspace embedding with random Khatri-Rao products and its application to eigensolvers

Accelerating the Computation of Tensor $Z$-eigenvalues