Primal-dual interior-point methods for semidefinite programming

Convergence rates, stability and numerical results

Farid Alizadeh, Jean Pierre A Haeberly, Michael L. Overton

Research output: Contribution to journalArticle

Abstract

Primal-dual interior-point path-following methods for semidefinite programming are considered. Several variants are discussed, based on Newton's method applied to three equations: primal feasibility, dual feasibility, and some form of centering condition. The focus is on three such algorithms, called the XZ, XZ+ZX, and Q methods. For the XZ+ZX and Q algorithms, the Newton system is well defined and its Jacobian is nonsingular at the solution, under nondegeneracy assumptions. The associated Schur complement matrix has an unbounded condition number on the central path under the nondegeneracy assumptions and an additional rank assumption. Practical aspects are discussed, including Mehrotra predictor-corrector variants and issues of numerical stability. Compared to the other methods considered, the XZ+ZX method is more robust with respect to its ability to step close to the boundary, converges more rapidly, and achieves higher accuracy.

Original languageEnglish (US)
Pages (from-to)746-768
Number of pages23
JournalSIAM Journal on Optimization
Volume8
Issue number3
StatePublished - Aug 1998

Fingerprint

Primal-dual Interior Point Method
Semidefinite Programming
Convergence Rate
Nondegeneracy
Numerical Results
Convergence of numerical methods
Newton-Raphson method
Path-following Methods
Central Path
Predictor-corrector
Schur Complement
Primal-dual
Interior Point
Numerical Stability
Condition number
Newton Methods
Well-defined
High Accuracy
Converge

Keywords

  • Convex programming
  • Eigenvalue optimization
  • Interior-point method
  • Semidefinite programming

ASJC Scopus subject areas

  • Mathematics(all)
  • Applied Mathematics

Cite this

Primal-dual interior-point methods for semidefinite programming : Convergence rates, stability and numerical results. / Alizadeh, Farid; Haeberly, Jean Pierre A; Overton, Michael L.

In: SIAM Journal on Optimization, Vol. 8, No. 3, 08.1998, p. 746-768.

Research output: Contribution to journalArticle

@article{51553719506d4940bc4d7f81d7ed6cd4,
title = "Primal-dual interior-point methods for semidefinite programming: Convergence rates, stability and numerical results",
abstract = "Primal-dual interior-point path-following methods for semidefinite programming are considered. Several variants are discussed, based on Newton's method applied to three equations: primal feasibility, dual feasibility, and some form of centering condition. The focus is on three such algorithms, called the XZ, XZ+ZX, and Q methods. For the XZ+ZX and Q algorithms, the Newton system is well defined and its Jacobian is nonsingular at the solution, under nondegeneracy assumptions. The associated Schur complement matrix has an unbounded condition number on the central path under the nondegeneracy assumptions and an additional rank assumption. Practical aspects are discussed, including Mehrotra predictor-corrector variants and issues of numerical stability. Compared to the other methods considered, the XZ+ZX method is more robust with respect to its ability to step close to the boundary, converges more rapidly, and achieves higher accuracy.",
keywords = "Convex programming, Eigenvalue optimization, Interior-point method, Semidefinite programming",
author = "Farid Alizadeh and Haeberly, {Jean Pierre A} and Overton, {Michael L.}",
year = "1998",
month = "8",
language = "English (US)",
volume = "8",
pages = "746--768",
journal = "SIAM Journal on Optimization",
issn = "1052-6234",
publisher = "Society for Industrial and Applied Mathematics Publications",
number = "3",

}

TY - JOUR

T1 - Primal-dual interior-point methods for semidefinite programming

T2 - Convergence rates, stability and numerical results

AU - Alizadeh, Farid

AU - Haeberly, Jean Pierre A

AU - Overton, Michael L.

PY - 1998/8

Y1 - 1998/8

N2 - Primal-dual interior-point path-following methods for semidefinite programming are considered. Several variants are discussed, based on Newton's method applied to three equations: primal feasibility, dual feasibility, and some form of centering condition. The focus is on three such algorithms, called the XZ, XZ+ZX, and Q methods. For the XZ+ZX and Q algorithms, the Newton system is well defined and its Jacobian is nonsingular at the solution, under nondegeneracy assumptions. The associated Schur complement matrix has an unbounded condition number on the central path under the nondegeneracy assumptions and an additional rank assumption. Practical aspects are discussed, including Mehrotra predictor-corrector variants and issues of numerical stability. Compared to the other methods considered, the XZ+ZX method is more robust with respect to its ability to step close to the boundary, converges more rapidly, and achieves higher accuracy.

AB - Primal-dual interior-point path-following methods for semidefinite programming are considered. Several variants are discussed, based on Newton's method applied to three equations: primal feasibility, dual feasibility, and some form of centering condition. The focus is on three such algorithms, called the XZ, XZ+ZX, and Q methods. For the XZ+ZX and Q algorithms, the Newton system is well defined and its Jacobian is nonsingular at the solution, under nondegeneracy assumptions. The associated Schur complement matrix has an unbounded condition number on the central path under the nondegeneracy assumptions and an additional rank assumption. Practical aspects are discussed, including Mehrotra predictor-corrector variants and issues of numerical stability. Compared to the other methods considered, the XZ+ZX method is more robust with respect to its ability to step close to the boundary, converges more rapidly, and achieves higher accuracy.

KW - Convex programming

KW - Eigenvalue optimization

KW - Interior-point method

KW - Semidefinite programming

UR - http://www.scopus.com/inward/record.url?scp=0032327612&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032327612&partnerID=8YFLogxK

M3 - Article

VL - 8

SP - 746

EP - 768

JO - SIAM Journal on Optimization

JF - SIAM Journal on Optimization

SN - 1052-6234

IS - 3

ER -