# Fundamentals of Matrix Computations

DAVID S. WATKINS
Washington State University

JOHN WILEY & SONS New York • Chichester • Brisbane • Toronto • Singapore

8 1.10 Introduction 86 Vector and Matrix Norms 87 Sensitivity of Linear Systems. Effects of Roundoff Errors 2.5 2.6 2.7 1.4 2.Contents 1 Gaussian Elimination and its Variants 1.3 1.2 2.9 2.8 2.9 1.7 2.2 1.6 1. Cholesky Decomposition 17 Banded Positive Definite Systems 33 Gaussian Elimination and the LU Decomposition 41 Gaussian Elimination with Pivoting 59 Banded Systems 69 The Influence of Computer Architecture on Algorithm Selection 73 1 2 Sensitivity of Linear Systems.3 2.1 1.5 1. Condition Numbers 94 Estimating the Condition Number 108 Analysis of Roundoff Errors 109 Gaussian Elimination with Ill-Conditioned Matrices 119 Why Small Pivots Should be Avoided 123 Limitations of the Norm Approach 127 Scaling 128 Iterative Refinement 132 86 xi .4 1.1 2.10 Systems of Linear Equations 1 Partitioning Matrices 4 The Problem 7 Triangulär Systems 9 Positive Definite Systems.

5 Sensitivity of Eigenvalues and Eigenvectors 325 5. Simultaneous Iteration. the Lanczos Algorithm 306 5.2 Subspace Iteration.4 Other Algorithms of QR Type 319 5.6 The Generalized Eigenvalue Problem 337 287 6 Other Method» for the Symmetrie Eigenvalue Problem 6.4 3.4 4.6 The Discrete Least-Squares Problem 134 Orthogonal Matrices. and Reflectors 136 Solution of the Least-Squares Problem 158 Orthonormal Vectors and the Gram-Schmidt Method 167 Geometrie Approach to the Least-Squares Problem 182 Sensitivity of the Least-Squares Problem 191 134 4 Eigenvalues and Eigenvectors I 4.3 Cuppen's Divide-and-Conquer Method 380 349 . and the QR Algorithm 293 5.5 3.1 Invariant Subspaces 287 5.Contents 3 Orthogonal Matrices and the Least-Squares Problem 3.3 Uniqueness Theorems for Hessenberg Form.6 4.2 The Slicing or Bisection Method 369 6. Rotators.1 The Jacobi Method 349 6.5 4.7 4.1 4.8 Systems of Differential Equations 199 Basic Properties of Eigenvalues and Eigenvectors 204 The Power Method and Some Simple Extensions 210 Similarity Transformations and Related Topics 224 Reduction to Hessenberg and Tridiagonal Forms 237 The QR Algorithm 251 Implementation of the QR Algorithm 265 Use of the QR Algorithm to Calculate Eigenvectors 279 199 5 Eigenvalues and Eigenvectors II 5.3 4.2 3.2 4.1 3.3 3.

4 7.Contents xiii 7 The Singular Value Decomposition 7.3 7.1 7.5 Introduction 390 Computing the SVD 398 Some Basic Applications of Singular Values 409 The SVD and the Least-Squares Problem 416 Angles and Distances between Subspaces 421 390 APPENDICES A Fast Botators B Software for Matrix Computations Bibliograph} Index 431 439 441 445 / .2 7.