NOX Solver AndersonAcceleration
NOX::Solver::AndersonAcceleration: Nonlinear solver based on Anderson Acceleration. NOX is short for Nonlinear Object-Oriented Solutions, and its objective is to enable the robust and efficient solution of the equation: $ F(x)=0 $, where $F:Re^n ightarrow Re^n$. NOX implements a variety of Newton-based globalization techniques including Line Search and Trust Region algorithms. In additon it provides higher and lower order models using Broyden and Tensor algorithms. Special algorithms have been developed for use with inexact linear solvers such as Krylov subspace techniques. NOX is designed to work with any linear algebra package and to be easily customized. NOX is part of Sandia’s Trilinos project.
References in zbMATH (referenced in 3 articles )
Showing results 1 to 3 of 3.
- Chen, Xiaojun; Kelley, C. T.: Convergence of the EDIIS algorithm for nonlinear equations (2019)
- Kelley, C. T.: Numerical methods for nonlinear equations (2018)
- Toth, Alex; Ellis, J. Austin; Evans, Tom; Hamilton, Steven; Kelley, C. T.; Pawlowski, Roger; Slattery, Stuart: Local improvement results for Anderson acceleration with inaccurate function evaluations (2017)