Convergence of the Nonmonotone Perry and Shanno Method for Optimization

Abstract

In this paper a new nonmonotone conjugate gradient method is introduced, which can be regarded as a generalization of the Perry and Shanno memoryless quasi-Newton method. For convex objective functions, the proposed nonmonotone conjugate gradient method is proved to be globally convergent. Its global convergence for non-convex objective functions has also been studied. Numerical experiments indicate that it is able to efficiently solve large scale optmization problems.

DOI: 10.1023/A:1008753308646

3 Figures and Tables

Cite this paper

@article{Liu2000ConvergenceOT, title={Convergence of the Nonmonotone Perry and Shanno Method for Optimization}, author={Guanghui Liu and Lili Jing}, journal={Comp. Opt. and Appl.}, year={2000}, volume={16}, pages={159-172} }