Abstract:We study theoretical properties of block solvers for the eigenvalue problem. Despite a recent surge of interest in such eigensolver analysis, truly block solvers have received relatively less attention, in contrast to the majority of studies concentrating on vector versions and non-truly block versions that rely on the deflation strategy. In fact, truly block solvers are more widely deployed in practice by virtue of its simplicity without compromise on accuracy. However, the corresponding theoretical analysis remains inadequate for first-order solvers, as only local and kth gap-dependent rates of convergence have been established thus far. This paper is devoted to revealing significantly better or asyet-unknown theoretical properties of such solvers. We present a novel convergence analysis in a unified framework for three types of first-order Riemannian solvers, i.e., deterministic, vanilla stochastic, and stochastic with variance reduction, that are to find top-k eigenvectors of a real symmetric matrix, in full generality. In particular, the issue of zero gaps between eigenvalues, to the best of our knowledge for the first time, is explicitly considered for these solvers, which brings new understandings, e.g., the dependence of convergence on gaps other than the k-th one. We thus propose the concept of generalized k-th gap. Three types of solvers are proved to converge to a globally optimal solution at a global, generalized k-th gap-dependent, and linear or sub-linear rate.

Inflationary dynamics for matrix eigenvalue problems

A Flexible Power Method for Solving Infinite Dimensional Tensor Eigenvalue Problems

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

The infinite Lanczos method for symmetric nonlinear eigenvalue problems

Exact Eigenvalues and Eigenvectors for Some n-Dimensional Matrices

Dynamical systems for eigenvalue problems of axisymmetric matrices with positive eigenvalues

A projected preconditioned conjugate gradient algorithm for computing many extreme eigenpairs of a Hermitian matrix

Dynamic Equations of Generalized Eigenvalue Problems

A Lanczos Method for Large-Scale Extreme Lorentz Eigenvalue Problems.

Eigenvalue programming beyond matrices

A Parallel Approach for Computing Complex Eigenvalue Problems

An Interesting Method For The Exponentials For Some Special Matrices

Matrix‐less methods for the spectral approximation of large non‐Hermitian Toeplitz matrices: A concise theoretical analysis and a numerical study

Matrix-less spectral approximation for large structured matrices

Large sparse symmetric eigenvalue problems with homogeneous linear constraints: the Lanczos process with inner–outer iterations

A numerical projection technique for large-scale eigenvalue problems

On joint eigen-decomposition of matrices

A Matrix Framework for the Solution of ODEs: Initial-, Boundary-, and Inner-Value Problems

A Krylov Eigenvalue Solver Based on Filtered Time Domain Solutions

Advances in Solution of Classical Generalized Eigenvalue Problem

On Truly Block Eigensolvers Via Riemannian Optimization.