DOCSLIB.ORG

AMS526: Numerical Analysis I (Numerical Linear Algebra for Computational and Data Sciences) Lecture 19: Computing the SVD; Sparse Linear Systems

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
AMS526: Numerical Analysis I (Numerical Linear Algebra for Computational and Data Sciences) Lecture 19: Computing the SVD; Sparse Linear Systems AMS526: Numerical Analysis I (Numerical Linear Algebra for Computational and Data Sciences) Lecture 19: Computing the SVD; Sparse Linear Systems Xiangmin Jiao Stony Brook University Xiangmin Jiao Numerical Analysis I 1 / 34 Outline 1 Computing the SVD (NLA§31) 2 Sparse Storage Format 3 Direct Methods for Sparse Linear Systems (MC§11.1-11.2) 4 Overview of Iterative Methods for Sparse Linear Systems Xiangmin Jiao Numerical Analysis I 2 / 34 SVD of A and Eigenvalues of A∗A m×n Intuitive idea for computing SVD of A 2 R : ∗ ∗ ∗ I Form A Apand computep its eigenvaluep p decomposition A A = V ΛV I Let Σ = Λ, i.e., diag( λ1; λ2;:::; λn) I Solve system UΣ = AV to obtain U This method is efficient if m n. However, it may not be stable, especially for smaller singular values because of the squaring of the condition number I For SVD of A, jσ~k − σk j = O(machinekAk), where σ~k and σk denote the computed and exact kth singular value ∗ I If computed from eigenvalue decomposition of A A, 2 jσ~k − σk j = O(machinekAk /σk ), which is problematic if σk kAk If one is interested in only relatively large singular values, then computing eigenvalues of A∗A is not a problem. For general situations, a more stable algorithm is desired. Xiangmin Jiao Numerical Analysis I 3 / 34 A Different Reduction to Eigenvalue Problem Typical algorithm for computing SVD are similar to computation of eigenvalues 0 A∗ Consider A 2 m×n, then Hermitian matrix H = has C A 0 eigenvalue decomposition VV VV Σ 0 H = ; U −U U −U 0 −Σ where A = UΣV ∗ gives the SVD. This approach is stable. In practice, such a reduction is done implicitly without forming the large matrix Typically done in two phases Xiangmin Jiao Numerical Analysis I 4 / 34 Two-Phase Method In the first phase, reduce to bidiagonal form by applying different orthogonal transformations on left and right, which involves O(mn2) operations In the second phase, reduce to diagonal form using a variant of QR algorithm or divide-and-conquer algorithm, which involves O(n2) operations for fixed precision We hereafter focus on the first phase Xiangmin Jiao Numerical Analysis I 5 / 34 Golub-Kahan Bidiagonalization Apply Householder reflectors on both left and right sides 2 4 3 Work for Golub-Kahan bidiagonalization ∼ 4mn − 3 n flops Xiangmin Jiao Numerical Analysis I 6 / 34 Lawson-Hanson-Chan Bidiagonalization Speed up by first performing QR factorization on A Work for LHC bidiagonalization ∼ 2mn2 + 2n3 flops, which is 5 advantageous if m ≥ 3 n Xiangmin Jiao Numerical Analysis I 7 / 34 Three-Step Bidiagonalization Hybrid approach: Apply QR at suitable time on submatrix with 5=3 aspect ratio 2 4 3 2 3 Work for three-step bidiagonalization is ∼ 4mn − 3 n − 3 (m − n) Xiangmin Jiao Numerical Analysis I 8 / 34 Comparison of Performance ) (Golub-Kahan One-step ) (LHC Two-step Three-step bidiagonalization Xiangmin Jiao Numerical Analysis I 9 / 34 Outline 1 Computing the SVD (NLA§31) 2 Sparse Storage Format 3 Direct Methods for Sparse Linear Systems (MC§11.1-11.2) 4 Overview of Iterative Methods for Sparse Linear Systems Xiangmin Jiao Numerical Analysis I 10 / 34 Sparse Linear System Boundary value problems and implicit methods for time-dependent PDEs yield systems of linear algebraic equations to solve A matrix is sparse if it has relatively few nonzeros in its entries Sparsity can be exploited to use far less than O(n2) storage and O(n3) work required in standard approach to solving system with dense matrix, assuming matrix is n × n Xiangmin Jiao Numerical Analysis I 11 / 34 Storage Format of Sparse Matrices Sparse-matrices are typically stored in special formats that store only nonzero entries, along with indices to identify their locations in matrix, such as I compressed-row storage (CRS) I compressed-column storage (CCS) I block compressed row storage (BCRS) Banded matrices have their own special storage formats (such as Compressed Diagonal Storage (CDS)) See survey at http://netlib.org/linalg/html_templates/node90.html Explicitly storing indices incurs additional storage overhead and makes arithmetic operations on nonzeros less efficient due to indirect addressing to access operands, so they are beneficial only for very sparse matrices Storage format can have big impact the effectiveness of different versions of same algorithm (with different ordering of loops) Besides direct methods, these storage formats are also important in implementing iterative and multigrid solvers Xiangmin Jiao Numerical Analysis I 12 / 34 Example of Compressed-Row Storage (CRS) Xiangmin Jiao Numerical Analysis I 13 / 34 Outline 1 Computing the SVD (NLA§31) 2 Sparse Storage Format 3 Direct Methods for Sparse Linear Systems (MC§11.1-11.2) 4 Overview of Iterative Methods for Sparse Linear Systems Xiangmin Jiao Numerical Analysis I 14 / 34 Banded Linear Systems Cost of factorizing banded linear system depends on bandwidth I For SPD n × n matrix with semi-bandwidth s, total flop count of Cholesky factorization is about ns2 I For n × n matrix with lower bandwidth p and upper bandwidth q, F In A = LU (LU without pivoting), total flop count is about 2npq F In PA = LU (LU with column pivoting), total flop count is about 2np(p + q) Banded matrices have their own special storage formats (such as Compressed Diagonal Storage (CDS)) Xiangmin Jiao Numerical Analysis I 15 / 34 Fill When applying LU or Cholesky factorization to general sparse matrix, taking linear combinations of rows or columns to annihilate unwanted nonzero entries can introduce new nonzeros into matrix locations that were initially zero Such new nonzeros, called fill or fill-in, must be stored and may themselves eventually need to be annihilated in order to obtain triangular factors Resulting triangular factors can be expected to contain at least as many nonzeros as original matrix and usually significant fill as well Xiangmin Jiao Numerical Analysis I 16 / 34 Sparse Cholesky Factorization In general, some heuristic algorithms are employed to reorder the matrix to reduce fills Amount of fill is sensitive to order in which rows and columns of matrix are processed, so basic problem in sparse factorization is reordering matrix to limit fill during factorization Exact minimization of fill is hard combinatorial problem (NP-complete), but heuristic algorithms such as minimum degree and nested dissection limit fill well for many types of problems For Cholesky factorization, both rows and columns are reordered Xiangmin Jiao Numerical Analysis I 17 / 34 Graph Model of Elimination Each step of factorization process corresponds to elimination of one node from graph Eliminating node causes its neighboring nodes to become connected to each other If any such neighbors were not already connected, then fill results (new edges in graph and new nonzeros in matrix) Commonly used reordering methods include Cuthill-McKee, approximate minimum degree ordering (AMD) and nested dissection Xiangmin Jiao Numerical Analysis I 18 / 34 Reordering to Reduce Bandwidth The Cuthill-McKee algorithm and reverse Cuthill-McKee algorithm I The Cuthill-McKee algorithm is a variant of the breadth-first search algorithm on graphs. F Starts with a peripheral node F Generates levels Ri for i = 1; 2;::: until all nodes are exhausted F The set Ri+1 is created from set Ri by listing all vertices adjacent to all nodes in Ri F Within each level, nodes are listed in increasing degree I The reverse Cuthill–McKee algorithm (RCM) reserves the resulting index numbers Xiangmin Jiao Numerical Analysis I 19 / 34 Approximate Minimum Degree Ordering Good heuristic for limiting fill is to eliminate first those nodes having fewest neighbors Number of neighbors is called degree of node, so heuristic is known as minimum degree At each step, select node of smallest degree for elimination, breaking ties arbitrarily After node has been eliminated, its neighbors become connected to each other, so degrees of some nodes may change Process is then repeated, with new node of minimum degree eliminated next, and so on until all nodes have been eliminated Xiangmin Jiao Numerical Analysis I 20 / 34 Minimum Degree Ordering, continued Cholesky factor suffers much less fill than with original ordering, and advantage grows with problem size Sophisticated versions of minimum degree are among most effective general-purpose orderings known Xiangmin Jiao Numerical Analysis I 21 / 34 Comparison of Different Orderings of Example Matrix Left: Nonzero pattern of matrix A. Right: Nonzero pattern of matrix R. Xiangmin Jiao Numerical Analysis I 22 / 34 Nested Dissection Ordering Nested dissection is based on divide-and-conquer First, small set of nodes is selected whose removal splits graph into two pieces of roughly equal size No node in either piece is connected to any node in other, so no fill occurs in either piece due to elimination of any node in the other Separator nodes are numbered last, then process is repeated recursively on each remaining piece of graph until all nodes have been numbered Xiangmin Jiao Numerical Analysis I 23 / 34 Nested Dissection Ordering Continued Dissection induces blocks of zeros in matrix that are automatically preserved during factorization Recursive nature of algorithm can be seen in hierarchical block structure of matrix, which would involve many more levels in larger problems Again, Cholesky factor suffers much less fill than with original ordering, and advantage grows with problem size Xiangmin Jiao Numerical Analysis I 24 / 34 Sparse Gaussian Elimination For Gaussian elimination, only columns are reordered Pivoting introduces additional fills in sparse Gaussian elimination Reordering may be done dynamically or statically The reverse Cuthill-McKee algorithm applied to A + AT may be used to reduce bandwidth Column approximate minimum-degree, may be employed to reorder matrix to reduce fills Xiangmin Jiao Numerical Analysis I 25 / 34 Comparison of Different Orderings of Example Matrix Nonzero pattern of A and L + U with random ordering.
Load more

Recommended publications
  • Fast Solution of Sparse Linear Systems with Adaptive Choice of Preconditioners Zakariae Jorti
    Fast solution of sparse linear systems with adaptive choice of preconditioners Zakariae Jorti To cite this version: Zakariae Jorti. Fast solution of sparse linear systems with adaptive choice of preconditioners. Gen- eral Mathematics [math.GM]. Sorbonne Université, 2019. English. NNT : 2019SORUS149. tel- 02425679v2 HAL Id: tel-02425679 https://tel.archives-ouvertes.fr/tel-02425679v2 Submitted on 15 Feb 2021 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Sorbonne Université École doctorale de Sciences Mathématiques de Paris Centre Laboratoire Jacques-Louis Lions Résolution rapide de systèmes linéaires creux avec choix adaptatif de préconditionneurs Par Zakariae Jorti Thèse de doctorat de Mathématiques appliquées Dirigée par Laura Grigori Co-supervisée par Ani Anciaux-Sedrakian et Soleiman Yousef Présentée et soutenue publiquement le 03/10/2019 Devant un jury composé de: M. Tromeur-Dervout Damien, Professeur, Université de Lyon, Rapporteur M. Schenk Olaf, Professeur, Università della Svizzera italiana, Rapporteur M. Hecht Frédéric, Professeur, Sorbonne Université, Président du jury Mme. Emad Nahid, Professeur, Université Paris-Saclay, Examinateur M. Vasseur Xavier, Ingénieur de recherche, ISAE-SUPAERO, Examinateur Mme. Grigori Laura, Directrice de recherche, Inria Paris, Directrice de thèse Mme. Anciaux-Sedrakian Ani, Ingénieur de recherche, IFPEN, Co-encadrante de thèse M.
    [Show full text]
  • Efficient “Black-Box” Multigrid Solvers for Convection-Dominated Problems
    EFFICIENT \BLACK-BOX" MULTIGRID SOLVERS FOR CONVECTION-DOMINATED PROBLEMS A thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Engineering and Physical Science 2011 Glyn Owen Rees School of Computer Science 2 Contents Declaration 12 Copyright 13 Acknowledgement 14 1 Introduction 15 2 The Convection-Di®usion Problem 18 2.1 The continuous problem . 18 2.2 The Galerkin approximation method . 22 2.2.1 The ¯nite element method . 26 2.3 The Petrov-Galerkin (SUPG) approximation . 29 2.4 Properties of the discrete operator . 34 2.5 The Navier-Stokes problem . 37 3 Methods for Solving Linear Algebraic Systems 42 3.1 Basic iterative methods . 43 3.1.1 The Jacobi method . 45 3.1.2 The Gauss-Seidel method . 46 3.1.3 Convergence of splitting iterations . 48 3.1.4 Incomplete LU (ILU) factorisation . 49 3.2 Krylov methods . 56 3.2.1 The GMRES method . 59 3.2.2 Preconditioned GMRES method . 62 3.3 Multigrid method . 67 3.3.1 Geometric multigrid method . 67 3.3.2 Algebraic multigrid method . 75 3.3.3 Parallel AMG . 80 3.3.4 Literature summary of multigrid . 81 3.3.5 Multigrid preconditioning of Krylov solvers . 84 3 3.4 A new tILU smoother . 85 4 Two-Dimensional Case Studies 89 4.1 The di®usion problem . 90 4.2 Geometric multigrid preconditioning . 95 4.2.1 Constant uni-directional wind . 95 4.2.2 Double glazing problem - recirculating wind . 106 4.2.3 Combined uni-directional and recirculating wind .
    [Show full text]
  • Chapter 7 Iterative Methods for Large Sparse Linear Systems
    Chapter 7 Iterative methods for large sparse linear systems In this chapter we revisit the problem of solving linear systems of equations, but now in the context of large sparse systems. The price to pay for the direct methods based on matrix factorization is that the factors of a sparse matrix may not be sparse, so that for large sparse systems the memory cost make direct methods too expensive, in memory and in execution time. Instead we introduce iterative methods, for which matrix sparsity is exploited to develop fast algorithms with a low memory footprint. 7.1 Sparse matrix algebra Large sparse matrices We say that the matrix A Rn is large if n is large, and that A is sparse if most of the elements are2 zero. If a matrix is not sparse, we say that the matrix is dense. Whereas for a dense matrix the number of nonzero elements is (n2), for a sparse matrix it is only (n), which has obvious implicationsO for the memory footprint and efficiencyO for algorithms that exploit the sparsity of a matrix. AdiagonalmatrixisasparsematrixA =(aij), for which aij =0for all i = j,andadiagonalmatrixcanbegeneralizedtoabanded matrix, 6 for which there exists a number p,thebandwidth,suchthataij =0forall i<j p or i>j+ p.Forexample,atridiagonal matrix A is a banded − 59 CHAPTER 7. ITERATIVE METHODS FOR LARGE SPARSE 60 LINEAR SYSTEMS matrix with p =1, xx0000 xxx000 20 xxx003 A = , (7.1) 600xxx07 6 7 6000xxx7 6 7 60000xx7 6 7 where x represents a nonzero4 element. 5 Compressed row storage The compressed row storage (CRS) format is a data structure for efficient represention of a sparse matrix by three arrays, containing the nonzero values, the respective column indices, and the extents of the rows.
    [Show full text]
  • Chebyshev and Fourier Spectral Methods 2000
    Chebyshev and Fourier Spectral Methods Second Edition John P. Boyd University of Michigan Ann Arbor, Michigan 48109-2143 email: [email protected] http://www-personal.engin.umich.edu/jpboyd/ 2000 DOVER Publications, Inc. 31 East 2nd Street Mineola, New York 11501 1 Dedication To Marilyn, Ian, and Emma “A computation is a temptation that should be resisted as long as possible.” — J. P. Boyd, paraphrasing T. S. Eliot i Contents PREFACE x Acknowledgments xiv Errata and Extended-Bibliography xvi 1 Introduction 1 1.1 Series expansions .................................. 1 1.2 First Example .................................... 2 1.3 Comparison with finite element methods .................... 4 1.4 Comparisons with Finite Differences ....................... 6 1.5 Parallel Computers ................................. 9 1.6 Choice of basis functions .............................. 9 1.7 Boundary conditions ................................ 10 1.8 Non-Interpolating and Pseudospectral ...................... 12 1.9 Nonlinearity ..................................... 13 1.10 Time-dependent problems ............................. 15 1.11 FAQ: Frequently Asked Questions ........................ 16 1.12 The Chrysalis .................................... 17 2 Chebyshev & Fourier Series 19 2.1 Introduction ..................................... 19 2.2 Fourier series .................................... 20 2.3 Orders of Convergence ............................... 25 2.4 Convergence Order ................................. 27 2.5 Assumption of Equal Errors ...........................
    [Show full text]
  • Relaxed Modulus-Based Matrix Splitting Methods for the Linear Complementarity Problem †
    S S symmetry Article Relaxed Modulus-Based Matrix Splitting Methods for the Linear Complementarity Problem † Shiliang Wu 1, *, Cuixia Li 1 and Praveen Agarwal 2,3 1 School of Mathematics, Yunnan Normal University, Kunming 650500, China; [email protected] 2 Department of Mathrematics, Anand International College of Engineering, Jaipur 303012, India; [email protected] 3 Nonlinear Dynamics Research Center (NDRC), Ajman University, Ajman, United Arab Emirates * Correspondence: [email protected] † This research was supported by National Natural Science Foundation of China (No.11961082). Abstract: In this paper, we obtain a new equivalent fixed-point form of the linear complementarity problem by introducing a relaxed matrix and establish a class of relaxed modulus-based matrix split- ting iteration methods for solving the linear complementarity problem. Some sufficient conditions for guaranteeing the convergence of relaxed modulus-based matrix splitting iteration methods are presented. Numerical examples are offered to show the efficacy of the proposed methods. Keywords: linear complementarity problem; matrix splitting; iteration method; convergence MSC: 90C33; 65F10; 65F50; 65G40 1. Introduction Citation: Wu, S.; Li, C.; Agarwal, P. Relaxed Modulus-Based Matrix In this paper, we focus on the iterative solution of the linear complementarity problem, Splitting Methods for the Linear abbreviated as ‘LCP(q, A)’, whose form is Complementarity Problem. Symmetry T 2021, 13, 503. https://doi.org/ w = Az + q ≥ 0, z ≥ 0 and z w = 0, (1) 10.3390/sym13030503 × where A 2 Rn n and q 2 Rn are given, and z 2 Rn is unknown, and for two s × t matrices Academic Editor: Jan Awrejcewicz G = (gij) and H = (hij) the order G ≥ (>)H means gij ≥ (>)hij for any i and j.
    [Show full text]
  • Numerical Solution of Saddle Point Problems
    Acta Numerica (2005), pp. 1–137 c Cambridge University Press, 2005 DOI: 10.1017/S0962492904000212 Printed in the United Kingdom Numerical solution of saddle point problems Michele Benzi∗ Department of Mathematics and Computer Science, Emory University, Atlanta, Georgia 30322, USA E-mail: [email protected] Gene H. Golub† Scientific Computing and Computational Mathematics Program, Stanford University, Stanford, California 94305-9025, USA E-mail: [email protected] J¨org Liesen‡ Institut f¨ur Mathematik, Technische Universit¨at Berlin, D-10623 Berlin, Germany E-mail: [email protected] We dedicate this paper to Gil Strang on the occasion of his 70th birthday Large linear systems of saddle point type arise in a wide variety of applica- tions throughout computational science and engineering. Due to their indef- initeness and often poor spectral properties, such linear systems represent a significant challenge for solver developers. In recent years there has been a surge of interest in saddle point problems, and numerous solution techniques have been proposed for this type of system. The aim of this paper is to present and discuss a large selection of solution methods for linear systems in saddle point form, with an emphasis on iterative methods for large and sparse problems. ∗ Supported in part by the National Science Foundation grant DMS-0207599. † Supported in part by the Department of Energy of the United States Government. ‡ Supported in part by the Emmy Noether Programm of the Deutsche Forschungs- gemeinschaft. 2 M. Benzi, G. H. Golub and J. Liesen CONTENTS 1 Introduction 2 2 Applications leading to saddle point problems 5 3 Properties of saddle point matrices 14 4 Overview of solution algorithms 29 5 Schur complement reduction 30 6 Null space methods 32 7 Coupled direct solvers 40 8 Stationary iterations 43 9 Krylov subspace methods 49 10 Preconditioners 59 11 Multilevel methods 96 12 Available software 105 13 Concluding remarks 107 References 109 1.
    [Show full text]
  • Copyright © by SIAM. Unauthorized Reproduction of This Article Is Prohibited. PARALLEL BIDIAGONALIZATION of a DENSE MATRIX 827
    SIAM J. MATRIX ANAL. APPL. c 2007 Society for Industrial and Applied Mathematics Vol. 29, No. 3, pp. 826–837 ∗ PARALLEL BIDIAGONALIZATION OF A DENSE MATRIX † ‡ ‡ § CARLOS CAMPOS , DAVID GUERRERO , VICENTE HERNANDEZ´ , AND RUI RALHA Abstract. A new stable method for the reduction of rectangular dense matrices to bidiagonal form has been proposed recently. This is a one-sided method since it can be entirely expressed in terms of operations with (full) columns of the matrix under transformation. The algorithm is well suited to parallel computing and, in order to make it even more attractive for distributed memory systems, we introduce a modification which halves the number of communication instances. In this paper we present such a modification. A block organization of the algorithm to use level 3 BLAS routines seems difficult and, at least for the moment, it relies upon level 2 BLAS routines. Nevertheless, we found that our sequential code is competitive with the LAPACK DGEBRD routine. We also compare the time taken by our parallel codes and the ScaLAPACK PDGEBRD routine. We investigated the best data distribution schemes for the different codes and we can state that our parallel codes are also competitive with the ScaLAPACK routine. Key words. bidiagonal reduction, parallel algorithms AMS subject classifications. 15A18, 65F30, 68W10 DOI. 10.1137/05062809X 1. Introduction. The problem of computing the singular value decomposition (SVD) of a matrix is one of the most important operations in numerical linear algebra and is employed in a variety of applications. The SVD is defined as follows. × For any rectangular matrix A ∈ Rm n (we will assume that m ≥ n), there exist ∈ Rm×m ∈ Rn×n ΣA ∈ Rm×n two orthogonal matrices U and V and a matrix Σ = 0 , t where ΣA = diag (σ1,...,σn) is a diagonal matrix, such that A = UΣV .
    [Show full text]
  • Trigonometric Transform Splitting Methods for Real Symmetric Toeplitz Systems
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Universidade do Minho: RepositoriUM TRIGONOMETRIC TRANSFORM SPLITTING METHODS FOR REAL SYMMETRIC TOEPLITZ SYSTEMS ZHONGYUN LIU∗, NIANCI WU∗, XIAORONG QIN∗, AND YULIN ZHANGy Abstract. In this paper we study efficient iterative methods for real symmetric Toeplitz systems based on the trigonometric transformation splitting (TTS) of the real symmetric Toeplitz matrix A. Theoretical analyses show that if the generating function f of the n × n Toeplitz matrix A is a real positive even function, then the TTS iterative methods converge to the unique solution of the linear system of equations for sufficient large n. Moreover, we derive an upper bound of the contraction factor of the TTS iteration which is dependent solely on the spectra of the two TTS matrices involved. Different from the CSCS iterative method in [19] in which all operations counts concern complex op- erations when the DFTs are employed, even if the Toeplitz matrix A is real and symmetric, our method only involves real arithmetics when the DCTs and DSTs are used. The numerical experiments show that our method works better than CSCS iterative method and much better than the positive definite and skew- symmetric splitting (PSS) iterative method in [3] and the symmetric Gauss-Seidel (SGS) iterative method. Key words. Sine transform, Cosine transform, Matrix splitting, Iterative methods, Real Toeplitz matrices. AMS subject classifications. 15A23, 65F10, 65F15. 1. Introduction. Consider the iterative solution to the following linear system of equa- tions Ax = b (1.1) by the two-step splitting iteration with alternation, where b 2 Rn and A 2 Rn×n is a symmetric positive definite Toeplitz matrix.
    [Show full text]
  • Numerical Linear Algebra
    Numerical Linear Algebra Comparison of Several Numerical Methods used to solve Poisson’s Equation Christiana Mavroyiakoumou University of Oxford A case study report submitted for the degree of M.Sc. in Mathematical Modelling and Scientific Computing Hilary 2017 1 Introduction In this report, we apply the finite difference scheme to the Poisson equation with homogeneous Dirichlet boundary conditions. This yields a system of linear equations with a large sparse system matrix that is a classical test problem for comparing direct and iterative linear solvers. The solution of sparse linear systems by iterative methods has become one of the core applications and research areas of scientific computing. The size of systems that are solved routinely has increased tremendously over time. This is because the discretisation of partial differential equations can lead to systems that are arbitrarily large. We introduce the reader to the general theory of regular splitting methods and present some of the classical methods such as Jacobi, Gauss-Seidel, Relaxed Jacobi, SOR and SSOR. Iterative methods yield the solution U of a linear system after an infinite number of steps. At each step, iterative methods require the computation of the residual of the system. In this report, we use iterative solution methods in order to solve n n n AU “ f;A P R ˆ ; f P R : (1.1) Linear systems can be solved by fixed point iteration. To do so, one must transform the system of linear equations into a fixed point form and this is in general achieved by splitting the matrix A into two parts, A “ M ´ N.
    [Show full text]
  • Minimum-Residual Methods for Sparse Least-Squares Using Golub-Kahan Bidiagonalization a Dissertation Submitted to the Institute
    MINIMUM-RESIDUAL METHODS FOR SPARSE LEAST-SQUARES USING GOLUB-KAHAN BIDIAGONALIZATION A DISSERTATION SUBMITTED TO THE INSTITUTE FOR COMPUTATIONAL AND MATHEMATICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY David Chin-lung Fong December 2011 © 2011 by Chin Lung Fong. All Rights Reserved. Re-distributed by Stanford University under license with the author. This dissertation is online at: http://purl.stanford.edu/sd504kj0427 ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Michael Saunders, Primary Adviser I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Margot Gerritsen I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Walter Murray Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives. iii iv ABSTRACT For 30 years, LSQR and its theoretical equivalent CGLS have been the standard iterative solvers for large rectangular systems Ax = b and least-squares problems min Ax b . They are analytically equivalent − to symmetric CG on the normal equation ATAx = ATb, and they reduce r monotonically, where r = b Ax is the k-th residual vector.
    [Show full text]
  • The Jacobi-Davidson Algorithm for Solving Large Sparse Symmetric Eigenvalue Problems with Application to the Design of Accelerator Cavities
    Research Collection Doctoral Thesis The Jacobi-Davidson algorithm for solving large sparse symmetric eigenvalue problems with application to the design of accelerator cavities Author(s): Geus, Roman Publication Date: 2002 Permanent Link: https://doi.org/10.3929/ethz-a-004469464 Rights / License: In Copyright - Non-Commercial Use Permitted This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use. ETH Library DISS. ETH NO. 14734 The Jacobi-Davidson algorithm for solving large sparse symmetric eigenvalue problems with application to the design of accelerator cavities A dissertation submitted to the SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH for the degree of Doctor of Technical Sciences presented by ROMAN GEUS Dipl. Informatik-Ing. ETH born 6. September 1970 citizen of Arbon, Switzerland accepted on the recommendation of Prof. Dr. Walter Gander, examiner Prof. Dr. Henk van der Vorst, co-examiner Dr. Peter Arbenz, co-examiner 2002 Acknowledgements This work would not have been possible without the help of many people. First of all I would like to thank Prof. Dr. Walter Gander for giving me the opportunity to do my Ph.D. thesis in his group. His guidance and encouragement were always a great help. My special thank goes to Dr. Peter Arbenz, my supervisor during all these years. With his support he contributed a great deal to this work. He was there to answer many of my questions and to help me dealing with difficult problems. He gave me the opportunity and freedom to learn a great variety of methods and to realise my own ideas.
    [Show full text]
  • Restarted Lanczos Bidiagonalization for the SVD in Slepc
    Scalable Library for Eigenvalue Problem Computations SLEPc Technical Report STR-8 Available at http://slepc.upv.es Restarted Lanczos Bidiagonalization for the SVD in SLEPc V. Hern´andez J. E. Rom´an A. Tom´as Last update: June, 2007 (slepc 2.3.3) Previous updates: { About SLEPc Technical Reports: These reports are part of the documentation of slepc, the Scalable Library for Eigenvalue Problem Computations. They are intended to complement the Users Guide by providing technical details that normal users typically do not need to know but may be of interest for more advanced users. Restarted Lanczos Bidiagonalization for the SVD in SLEPc STR-8 Contents 1 Introduction 2 2 Description of the Method3 2.1 Lanczos Bidiagonalization...............................3 2.2 Dealing with Loss of Orthogonality..........................6 2.3 Restarted Bidiagonalization..............................8 2.4 Available Implementations............................... 11 3 The slepc Implementation 12 3.1 The Algorithm..................................... 12 3.2 User Options...................................... 13 3.3 Known Issues and Applicability............................ 14 1 Introduction The singular value decomposition (SVD) of an m × n complex matrix A can be written as A = UΣV ∗; (1) ∗ where U = [u1; : : : ; um] is an m × m unitary matrix (U U = I), V = [v1; : : : ; vn] is an n × n unitary matrix (V ∗V = I), and Σ is an m × n diagonal matrix with nonnegative real diagonal entries Σii = σi for i = 1;:::; minfm; ng. If A is real, U and V are real and orthogonal. The vectors ui are called the left singular vectors, the vi are the right singular vectors, and the σi are the singular values. In this report, we will assume without loss of generality that m ≥ n.
    [Show full text]