Error Analysis of Krylov Subspace approximation Based on IDR($s$) Method for Matrix Function Bilinear Forms

Abstract

The matrix function bilinear form, namely $\mathbf{u}^\top f(A)\mathbf{v}$, appears in many scientific computing problems, where $\mathbf{u}, \mathbf{v} \in \mathbb{R}^n$, $A \in \mathbb{R}^{n\times n}$, and $f(z)$ is a given analytic function. The Induced Dimension Reduction (IDR($s$)) method was originally proposed to solve a large system of linear equations, and effectively reduces the complexity and storage requirement by dimensionality reduction while maintaining the numerical stability of the algorithm. In fact, the IDR($s$) method can generate an interesting Hessenberg decomposition. We make use of this decomposition to establish the numerical algorithm and a practical error estimation framework for the matrix function bilinear form. Based on an error analysis of the IDR($s$) approximation, the corresponding error expansion is derived. Crucially, we prove that, under mild conditions, the remainder term in this expansion converges to zero as the truncation order increases, thereby validating the error expansion. The leading computable term is then used as a practical a posteriori error indicator for general analytic functions. We present numerical experiments to support our theoretical findings and illustrate the efficacy of our proposed method, along with its stopping criterion, over traditional Bi-Lanczos and Arnoldi-based algorithms.