Algorithm 851: CG_DESCENT, a conjugate gradient method with guaranteed descent

William W. Hager, Hongchao Zhang

Code and Data Abstract

Recently, a new nonlinear conjugate gradient scheme was developed which satisfies the descent condition gTkdk ≤ −7/8 ‖gk‖2 and which is globally convergent whenever the line search fulfills the Wolfe conditions. This article studies the convergence behavior of the algorithm; extensive numerical tests and comparisons with other methods for large-scale unconstrained optimization are given.

Article

Paper Abstract

Recently, a new nonlinear conjugate gradient scheme was developed which satisfies the descent condition gTkdk ≤ −7/8 ‖gk‖2 and which is globally convergent whenever the line search fulfills the Wolfe conditions. This article studies the convergence behavior of the algorithm; extensive numerical tests and comparisons with other methods for large-scale unconstrained optimization are given.

William W. Hager, Hongchao Zhang, et al. " Algorithm 851: CG_DESCENT, a conjugate gradient method with guaranteed descent." ACM Transactions on Mathematical Software.     doi:10.1145/1132973.1132979. Retrieved 10/23/2018 from researchcompendia.org/compendia/2013.290/

Compendium Type: Published Papers
Primary Research Field: Computer and Information Sciences
Secondary Research Field: Mathematics
Content License: Public Domain Mark
Code License: MIT License

Page Owner

jenn.seiler@gmail.com

created 12/12/2013

modified 01/16/2014

blog comments powered by Disqus