Stability implies robust convergence of a class of diagonalization-based iterative algorithms

02/09/2021 ∙ by Shulin Wu, et al. ∙ 0

Solving wave equations in a time-parallel manner is challenging, and the algorithm based on the block α-circulant preconditioning technique has shown its advantage in many existing studies (where α∈(0, 1) is a free parameter). Considerable efforts have been devoted to exploring the spectral radius of the preconditioned matrix and this leads to many case-by-case studies depending on the used time-integrator. In this paper, we propose a unified way to analyze the convergence via directly studying the error of the algorithm and using the stability of the time integrator. Our analysis works for all one-step time-integrators and two exemplary classes of two-step time-integrators: the parameterized Numerov methods and the parameterized two-stage hybrid methods. The main conclusion is that the global error satisfies err^k+1_𝐖, 2≤α/1-α𝐞𝐫𝐫^k_𝐖, 2 provided that the time-integrator is stable, where k is the iteration index and for any vector v the norm v_𝐖, 2 is defined by v_𝐖, 2=𝐖v_2 with 𝐖 being a matrix depending on the space discretization matrix only. Even though we focus on wave equations in this paper, the proposed argument is directly applicable to other evolution problems.



There are no comments yet.


page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.