Regularization methods for semidefinite programming

Jerome Malick (jerome.malickinria.fr)
Janez Povh (janez.povhuni-mb.si)
Franz Rendl (franz.rendluni-klu.ac.at)
Angelika Wiegele (angelika.wiegeleuni-klu.ac.at)

Abstract: This paper studies an alternative technique to interior point methods and nonlinear methods for semidefinite programming (SDP). The approach based on classical quadratic regularizations leads to an algorithm, generalizing a recent method called "boundary point method". We study the theoretical properties of this algorithm and we show that in practice it behaves very well on some instances of SDP having a large number of constraints.

Keywords: semidefinite programming, regularization methods, Augmented Lagrangian method, large scale semidefinite problems

Category 1: Convex and Nonsmooth Optimization (Convex Optimization )

Category 2: Linear, Cone and Semidefinite Programming (Semi-definite Programming )

Category 3: Nonlinear Optimization

Citation: Technical Report, Alpen-Adria-Universität Klagenfurt, Austria, October 2007.

Download: [Postscript][PDF]

Entry Submitted: 10/05/2007
Entry Accepted: 10/05/2007
Entry Last Modified: 10/11/2007

Modify/Update this entry