Abstract:We consider the solution of large stiff systems of ordinary differential equations with explicit exponential Runge--Kutta integrators. These problems arise from semi-discretized semi-linear parabolic partial differential equations on continuous domains or on inherently discrete graph domains. A series of results reduces the requirement of computing linear combinations of $\varphi$-functions in exponential integrators to the approximation of the action of a smaller number of matrix exponentials on certain vectors. State-of-the-art computational methods use polynomial Krylov subspaces of adaptive size for this task. They have the drawback that the required number of Krylov subspace iterations to obtain a desired tolerance increase drastically with the spectral radius of the discrete linear differential operator, e.g., the problem size. We present an approach that leverages rational Krylov subspace methods promising superior approximation qualities. We prove a novel a-posteriori error estimate of rational Krylov approximations to the action of the matrix exponential on vectors for single time points, which allows for an adaptive approach similar to existing polynomial Krylov techniques. We discuss pole selection and the efficient solution of the arising sequences of shifted linear systems by direct and preconditioned iterative solvers. Numerical experiments show that our method outperforms the state of the art for sufficiently large spectral radii of the discrete linear differential operators. The key to this are approximately constant numbers of rational Krylov iterations, which enable a near-linear scaling of the runtime with respect to the problem size.

An Arnoldi-Extrapolation algorithm for computing PageRank

Accelerating the Arnoldi-Type Algorithm for the PageRank Problem and the ProteinRank Problem

A Power–Arnoldi Algorithm for Computing PageRank

Arnoldi Versus GMRES for Computing Pagerank

Enhanced preprocessed multi-step splitting iterations for computing PageRank

Shifted power-GMRES method accelerated by extrapolation for solving PageRank with multiple damping factors

Improving eigenvectors in Arnoldi's method

A VARIANT OF THE REFINED ARNOLDI METHOD FOR LARGE SCALE MATRIX EIGENPROBLEMS

Probabilistic Richardson Extrapolation

Polynomial characterizations of the approximate eigenvectors by the refined Arnoldi method and an implicitly restarted refined Arnoldi algorithm

The Arnoldi Process, Short Recursions, and Displacement Ranks

Weak dangling block reordering and multi-step block compression for efficiently computing and updating PageRank solutions

Randomized flexible GMRES with deflated restarting

A Block Arnoldi Algorithm Based Reduced-Order Model Applied to Large-Scale Algebraic Equations of a 3-D Field Problem

An Inexact Shift‐and‐invert Arnoldi Algorithm for Toeplitz Matrix Exponential

Accurate Polynomial Fitting and Evaluation Via Arnoldi

Adaptive rational Krylov methods for exponential Runge--Kutta integrators

A Refined Harmonic Rayleigh-Ritz Procedure and an Explicitly Restarted Refined Harmonic Arnoldi Algorithm

On a family of low-rank algorithms for large-scale algebraic Riccati equations

Krylov subspace algorithms for computing GeneRank for the analysis of microarray data mining.

A Variation on the Block Arnoldi Method for Large Unsymmetric Matrix Eigenproblems