Perturbation Theory for Factorizations of LU Type through Series Expansions

Component- and normwise perturbation bounds for the block LU factorization and block LDL$^*$ factorization of Hermitian matrices are presented. We also obtain, as a consequence, perturbation bounds for the usual pointwise LU, LDL$^*$, and Cholesky factorizations. Some of these latter bounds are already known, but others improve previous results. All the bounds presented are easily proved by using series expansions. Given a square matrix $A=L U$ having the LU factorization, and a perturbation $E$, the LU factors of the matrix $A+E= \widetilde{L} \widetilde{U}$ are written as two convergent series of matrices: $\widetilde{L} = \sum_{k=0}^{\infty} L_k$ and $\widetilde{U} = \sum_{k=0}^{\infty} U_k$, where $L_k = O(\|E\|^k)$, $U_k = O(\|E\|^k)$, and $L_0 = L$, $U_0 = U$. We present expressions for the matrices $L_k$ and $U_k$ in terms of $L$, $U$, and $E$. The domain and the rate of convergence of these series are studied. Simple bounds on the remainders of any order of these series are found, which significantly improve the bounds on the second-order terms existing in the literature. This is useful when first-order perturbation analysis is used.

[1]  Alan Edelman,et al.  On Parlett's matrix norm inequality for the Cholesky decomposition , 1995, Numer. Linear Algebra Appl..

[2]  John G. Lewis,et al.  Accurate Symmetric Indefinite Linear Equation Solvers , 1999, SIAM J. Matrix Anal. Appl..

[3]  Ji-guang Sun Perturbation bounds for the Cholesky andQR factorizations , 1991 .

[4]  Xiao-Wen Chang Some Features of Gaussian Elimination with Rook Pivoting , 2002 .

[5]  Xiao-Wen Chang,et al.  On the sensitivity of the LU factorization , 1998 .

[6]  A. Barrlund Perturbation bounds for theLDLH andLU decompositions , 1991 .

[7]  J. Demmel,et al.  Computing the Singular Value Decomposition with High Relative Accuracy , 1997 .

[8]  Beresford N. Parlett,et al.  The New qd Algorithms , 1995, Acta Numerica.

[9]  Xiao-Wen Chang,et al.  Sensitivity analyses for factorizations of sparse or structured matrices , 1998 .

[10]  A. Sluis Condition numbers and equilibration of matrices , 1969 .

[11]  Ed Anderson,et al.  LAPACK Users' Guide , 1995 .

[12]  J. Bunch,et al.  Some stable methods for calculating inertia and solving symmetric linear systems , 1977 .

[13]  J. Bunch,et al.  Direct Methods for Solving Symmetric Indefinite Systems of Linear Equations , 1971 .

[14]  G. W. Stewarty On the Perturbation of LU, Cholesky, and QR Factorizations , 2007 .

[15]  Ji-guang Sun,et al.  Rounding-error and perturbation bounds for the Cholesky and LDL T factorizations , 1991 .

[16]  Gene H. Golub,et al.  Matrix computations (3rd ed.) , 1996 .

[17]  G. Stewart On the perturbation of LU and Cholesky factors , 1997 .

[18]  Nicholas J. Higham,et al.  INVERSE PROBLEMS NEWSLETTER , 1991 .

[19]  Ji-guang Sun Componentwise perturbation bounds for some matrix decompositions , 1992 .

[20]  Kenneth H. Rosen,et al.  Discrete Mathematics and its applications , 2000 .

[21]  Charles R. Johnson,et al.  Matrix analysis , 1985, Statistical Inference for Engineers and Data Scientists.

[22]  Z. Drmač,et al.  On the Perturbation of the Cholesky Factorization , 1994 .

[23]  James Demmel,et al.  Stability of block LU factorization , 1992, Numer. Linear Algebra Appl..

[24]  M. SIAMJ. STABILITY OF THE DIAGONAL PIVOTING METHOD WITH PARTIAL PIVOTING , 1995 .

[25]  Stability and Sensitivity of Tridiagonal LU Factorization without Pivoting , 2004 .

[26]  G. Stewart Perturbation Bounds for the $QR$ Factorization of a Matrix , 1977 .

[27]  Ivan Slapničar,et al.  Componentwise Analysis of Direct Factorization of Real Symmetric and Hermitian Matrices , 1998 .

[28]  A. Shapiro Optimal Block Diagonal $l_2 $-Scaling of Matrices , 1985 .

[29]  G. Stewart,et al.  New perturbation analyses for the Cholesky factorization , 1996 .