Properties of a representation of a basis for the null space

Philip E. Gill, Walter Murray, Michael A. Saunders, G. W. Stewart, Margaret Wright

Research output: Contribution to journalArticle

Abstract

Given a rectangular matrix A(x) that depends on the independent variables x, many constrained optimization methods involve computations with Z(x), a matrix whose columns form a basis for the null space of AT(x). When A is evaluated at a given point, it is well known that a suitable Z (satisfying ATZ = 0) can be obtained from standard matrix factorizations. However, Coleman and Sorensen have recently shown that standard orthogonal factorization methods may produce orthogonal bases that do not vary continuously with x; they also suggest several techniques for adapting these schemes so as to ensure continuity of Z in the neighborhood of a given point. This paper is an extension of an earlier note that defines the procedure for computing Z. Here, we first describe how Z can be obtained by updating an explicit QR factorization with Householder transformations. The properties of this representation of Z with respect to perturbations in A are discussed, including explicit bounds on the change in Z. We then introduce regularized Householder transformations, and show that their use implies continuity of the full matrix Q. The convergence of Z and Q under appropriate assumptions is then proved. Finally, we indicate why the chosen form of Z is convenient in certain methods for nonlinearly constrained optimization.

Original languageEnglish (US)
Pages (from-to)172-186
Number of pages15
JournalMathematical Programming
Volume33
Issue number2
DOIs
StatePublished - Nov 1985

Fingerprint

Householder Transformation
Null Space
Constrained Optimization
Factorization
Orthogonal Factorization
QR Factorization
Q-matrix
Orthogonal Basis
Factorization Method
Matrix Factorization
Explicit Bounds
Constrained optimization
Updating
Optimization Methods
Vary
Perturbation
Imply
Computing
Standards
Continuity

Keywords

  • Matrix Factorization
  • Nonlinear Optimization
  • Null-Space Continuity

ASJC Scopus subject areas

  • Applied Mathematics
  • Mathematics(all)
  • Safety, Risk, Reliability and Quality
  • Management Science and Operations Research
  • Software
  • Computer Graphics and Computer-Aided Design
  • Computer Science(all)
  • Engineering(all)

Cite this

Properties of a representation of a basis for the null space. / Gill, Philip E.; Murray, Walter; Saunders, Michael A.; Stewart, G. W.; Wright, Margaret.

In: Mathematical Programming, Vol. 33, No. 2, 11.1985, p. 172-186.

Research output: Contribution to journalArticle

Gill, PE, Murray, W, Saunders, MA, Stewart, GW & Wright, M 1985, 'Properties of a representation of a basis for the null space', Mathematical Programming, vol. 33, no. 2, pp. 172-186. https://doi.org/10.1007/BF01582244
Gill, Philip E. ; Murray, Walter ; Saunders, Michael A. ; Stewart, G. W. ; Wright, Margaret. / Properties of a representation of a basis for the null space. In: Mathematical Programming. 1985 ; Vol. 33, No. 2. pp. 172-186.
@article{c4f58ebf8e6c45b79e6f701e5e198651,
title = "Properties of a representation of a basis for the null space",
abstract = "Given a rectangular matrix A(x) that depends on the independent variables x, many constrained optimization methods involve computations with Z(x), a matrix whose columns form a basis for the null space of AT(x). When A is evaluated at a given point, it is well known that a suitable Z (satisfying ATZ = 0) can be obtained from standard matrix factorizations. However, Coleman and Sorensen have recently shown that standard orthogonal factorization methods may produce orthogonal bases that do not vary continuously with x; they also suggest several techniques for adapting these schemes so as to ensure continuity of Z in the neighborhood of a given point. This paper is an extension of an earlier note that defines the procedure for computing Z. Here, we first describe how Z can be obtained by updating an explicit QR factorization with Householder transformations. The properties of this representation of Z with respect to perturbations in A are discussed, including explicit bounds on the change in Z. We then introduce regularized Householder transformations, and show that their use implies continuity of the full matrix Q. The convergence of Z and Q under appropriate assumptions is then proved. Finally, we indicate why the chosen form of Z is convenient in certain methods for nonlinearly constrained optimization.",
keywords = "Matrix Factorization, Nonlinear Optimization, Null-Space Continuity",
author = "Gill, {Philip E.} and Walter Murray and Saunders, {Michael A.} and Stewart, {G. W.} and Margaret Wright",
year = "1985",
month = "11",
doi = "10.1007/BF01582244",
language = "English (US)",
volume = "33",
pages = "172--186",
journal = "Mathematical Programming",
issn = "0025-5610",
publisher = "Springer-Verlag GmbH and Co. KG",
number = "2",

}

TY - JOUR

T1 - Properties of a representation of a basis for the null space

AU - Gill, Philip E.

AU - Murray, Walter

AU - Saunders, Michael A.

AU - Stewart, G. W.

AU - Wright, Margaret

PY - 1985/11

Y1 - 1985/11

N2 - Given a rectangular matrix A(x) that depends on the independent variables x, many constrained optimization methods involve computations with Z(x), a matrix whose columns form a basis for the null space of AT(x). When A is evaluated at a given point, it is well known that a suitable Z (satisfying ATZ = 0) can be obtained from standard matrix factorizations. However, Coleman and Sorensen have recently shown that standard orthogonal factorization methods may produce orthogonal bases that do not vary continuously with x; they also suggest several techniques for adapting these schemes so as to ensure continuity of Z in the neighborhood of a given point. This paper is an extension of an earlier note that defines the procedure for computing Z. Here, we first describe how Z can be obtained by updating an explicit QR factorization with Householder transformations. The properties of this representation of Z with respect to perturbations in A are discussed, including explicit bounds on the change in Z. We then introduce regularized Householder transformations, and show that their use implies continuity of the full matrix Q. The convergence of Z and Q under appropriate assumptions is then proved. Finally, we indicate why the chosen form of Z is convenient in certain methods for nonlinearly constrained optimization.

AB - Given a rectangular matrix A(x) that depends on the independent variables x, many constrained optimization methods involve computations with Z(x), a matrix whose columns form a basis for the null space of AT(x). When A is evaluated at a given point, it is well known that a suitable Z (satisfying ATZ = 0) can be obtained from standard matrix factorizations. However, Coleman and Sorensen have recently shown that standard orthogonal factorization methods may produce orthogonal bases that do not vary continuously with x; they also suggest several techniques for adapting these schemes so as to ensure continuity of Z in the neighborhood of a given point. This paper is an extension of an earlier note that defines the procedure for computing Z. Here, we first describe how Z can be obtained by updating an explicit QR factorization with Householder transformations. The properties of this representation of Z with respect to perturbations in A are discussed, including explicit bounds on the change in Z. We then introduce regularized Householder transformations, and show that their use implies continuity of the full matrix Q. The convergence of Z and Q under appropriate assumptions is then proved. Finally, we indicate why the chosen form of Z is convenient in certain methods for nonlinearly constrained optimization.

KW - Matrix Factorization

KW - Nonlinear Optimization

KW - Null-Space Continuity

UR - http://www.scopus.com/inward/record.url?scp=0022162615&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0022162615&partnerID=8YFLogxK

U2 - 10.1007/BF01582244

DO - 10.1007/BF01582244

M3 - Article

VL - 33

SP - 172

EP - 186

JO - Mathematical Programming

JF - Mathematical Programming

SN - 0025-5610

IS - 2

ER -