Convergence of the EDIIS algorithm for nonlinear equations

Xiaojun Chen, C. T. Kelley

Research output: Journal article publicationJournal articleAcademic researchpeer-review

2 Citations (Scopus)

Abstract

The Energy Direct Inversion on the Iterative Subspace (EDIIS) algorithm was designed to globalize Anderson acceleration, a method for improving the performance of fixed point iteration. The motivating application is electronic structure computations. In this paper we prove a convergence result for that algorithm and illustrate the theory with a computational example.

Original languageEnglish
Pages (from-to)A365-A379
Number of pages15
JournalSIAM Journal on Scientific Computing
Volume41
Issue number1
DOIs
Publication statusPublished - 17 Jan 2019

Keywords

  • Anderson acceleration
  • EDIIS
  • Nonlinear equations

ASJC Scopus subject areas

  • Computational Mathematics
  • Applied Mathematics

Cite this