Gradient Methods for Large Scale Convex Quadratic Functions

Yaxiang Yuan ()
Additional contact information
Yaxiang Yuan: Chinese Academy of Sciences, State Key Laboratory of Scientific/Engineering Computing, Institute of Computational Mathematics and Scientific/Engineering Computing, Academy of Mathematics and Systems Science

Chapter Chapter 7 in Optimization and Regularization for Computational Inverse Problems and Applications, 2010, pp 141-155 from Springer

Abstract: Abstract The gradient method is one of the most simple methods for solving unconstrained optimization, it has the advantages of being easy to program and suitable for large scale problems. Different step-lengths give different gradient algorithms. In 1988, Barzilai and Borwein gave two interesting choices for the step-length and established superlinearly convergence results for two-dimensional convex quadratic problems. Barzilai and Borwein’s work triggered much research on the gradient method in the past two decades. In this chapter we investigate how the BB method can be further improved. We generalize the convergence result for the gradient method with retards. Our generalization allows more choices for the step-lengths. An intuitive analysis is given on the impact of the step-length for the speed of convergence of the gradient method. We propose a short BB step-length method. Numerical results on random generated problems are given to show that our short step technique can improve the BB method for large scale and ill-conditioned problems, particularly when high accurate solutions are needed.

Keywords: Gradient Method; Large Scale Problem; Steep Descent Method; Superlinear Convergence; Trust Region Method (search for similar items in EconPapers)
Date: 2010
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:sprchp:978-3-642-13742-6_7

Ordering information: This item can be ordered from
http://www.springer.com/9783642137426

DOI: 10.1007/978-3-642-13742-6_7

Access Statistics for this chapter

More chapters in Springer Books from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().