On condition numbers for least squares with quadric inequality constraint

被引：10

作者：

Diao, Huai-An ^{[1
]}

机构：

[1] Northeast Normal Univ, Sch Math & Stat, Changchun 130024, Peoples R China

来源：

COMPUTERS & MATHEMATICS WITH APPLICATIONS | 2017年 / 73卷 / 04期

关键词：

General least squares with quadric inequality constraint; Condition number; Componentwise; Perturbations; COMPONENTWISE CONDITION NUMBERS; STRUCTURED CONDITION NUMBERS; MOORE-PENROSE INVERSE; TIKHONOV REGULARIZATION; MATRICES;

D O I：

10.1016/j.camwa.2016.12.033

中图分类号：

O29 [应用数学];

学科分类号：

070104 ;

摘要：

In this paper, we will study normwise, mixed and componentwise condition numbers for the linear mapping of the solution for general least squares with quadric inequality constraint (GLSQI) and its standard form (LSQI). We will introduce the mappings from the data space to the interested data space, and the Frechet derivative of the introduced mapping can deduced through matrix differential techniques. Based on condition number theory, we derive the explicit expressions of normwise, mixed and componentwise condition numbers for the linear function of the solution for GLSQI and LSQI. Also, easier computable upper bounds for mixed and componentwise condition numbers are given. Numerical example shows that the mixed and componentwise condition numbers can tell us the true conditioning of the problem when its data is sparse or badly scaled. Compared with normwise condition numbers, the mixed and componentwise condition number can give sharp perturbation bounds. (C) 2017 Elsevier Ltd. All rights reserved.

引用

页码：616 / 627

页数：12

共 28 条

[1]

ANDERSON E., 1999, LAPACK USERSGUIDE, V3rd

[2]

[Anonymous], 2002, Accuracy and stability of numerical algorithms

[3]

[Anonymous], 2001, Matrix Analysis and Applied Linear Algebra

[4] A partial condition number for linear least squares problems [J].

Arioli, Mario ;

Baboulin, Marc ;

Gratton, Serge .

SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2007, 29 (02) :413-433

[5] Using dual techniques to derive componentwise and mixed condition numbers for a linear function of a linear least squares solution [J].

Baboulin, Marc ;

Gratton, Serge .

BIT NUMERICAL MATHEMATICS, 2009, 49 (01) :3-19

[6]

Ben-Israel A., 2003, Generalized inverses: theory and applications, V15

[7]

Bjorck A, 1996, NUMERICAL METHODS LE

[8]

Burgisser P., 2013, Condition: The Geometry of Numerical Algorithms

[9] Condition numbers and perturbation analysis for the Tikhonov regularization of discrete ill-posed problems [J].

Chu, Delin ;

Lin, Lijing ;

Tan, Roger C. E. ;

Wei, Yimin .

NUMERICAL LINEAR ALGEBRA WITH APPLICATIONS, 2011, 18 (01) :87-103

[10] Mixed and componentwise condition numbers for rectangular structured matrices [J].

Cucker, Felipe ;

Diao, Huaian .

CALCOLO, 2007, 44 (02) :89-115

← 1 2 3 →