Practical shift choice in the shift-and-invert Krylov subspace evaluations of the matrix exponential

TL;DR

This paper introduces two novel methods for selecting optimal shift parameters in the shift-and-invert Krylov subspace technique, improving the computation of matrix exponential-vector products especially when handling multiple vectors.

Contribution

It presents two new approaches for adaptive shift selection in the shift-and-invert method, enhancing efficiency in matrix exponential computations for multiple vectors.

Findings

Methods outperform fixed shift strategies in numerical experiments.

Proposed techniques effectively handle complex PDEs with discontinuities.

Adaptive shift selection reduces computational cost and improves accuracy.

Abstract

We propose two methods to find a proper shift parameter in the shift-and-invert method for computing matrix exponential matrix-vector products. These methods are useful in the case of matrix exponential action has to be computed for a number of vectors. The first method is based on the zero-order optimization of the mean residual norm for a given number of initial vectors. The second method processes the vectors one-by-one and estimates, for each vector, the derivative of the residual norm as a function of the shift parameter. The estimated derivative value is then used to update the shift value for the next vector. To demonstrate the performance of the proposed methods we perform numerical experiments for two-dimensional non-stationary convection-diffusion equation with discontinuous coefficients and two-dimensional anisotropic diffusion equation.