TITLE

Preconditioned Low-Rank Methods for High-Dimensional Elliptic PDE Eigenvalue Problems

AUTHOR(S)
Kressner, Daniel; Tobler, Christine
PUB. DATE
July 2011
SOURCE
Computational Methods in Applied Mathematics;2011, Vol. 11 Issue 3, p363
SOURCE TYPE
Academic Journal
DOC. TYPE
Article
ABSTRACT
We consider elliptic PDE eigenvalue problems on a tensorized domain, discretized such that the resulting matrix eigenvalue problem Ax = λx exhibits Kronecker product structure. In particular, we are concerned with the case of high dimensions, where standard approaches to the solution of matrix eigenvalue problems fail due to the exponentially growing degrees of freedom. Recent work shows that this curse of dimensionality can in many cases be addressed by approximating the desired solution vector x in a low-rank tensor format. In this paper, we use the hierarchical Tucker decomposition to develop a low-rank variant of LOBPCG, a classical preconditioned eigenvalue solver. We also show how the ALS and MALS (DMRG) methods known from computational quantum physics can be adapted to the hierarchical Tucker decomposition. Finally, a combination of ALS and MALS with LOBPCG and with our low-rank variant is proposed. A number of numerical experiments indicate that such combinations represent the methods of choice.
ACCESSION #
69616501

 

Related Articles

  • SINE TRANSFORM MATRIX FOR SOLVING TOEPLITZ MATRIX PROBLEMS. Li-zhi Cheng // Journal of Computational Mathematics;Mar2001, Vol. 19 Issue 2, p167 

    Provides information on a study which discussed the properties of eigenvalues for the solutions of symmetric positive definite Toeplitz systems, skew circulant and sine transform based properties. Eigenvalues of various preconditioners; Design of positive sine transform based preconditioners;...

  • A New Gradient Method with an Optimal Stepsize Property. Dai, Y. H.; Yang, X. Q. // Computational Optimization & Applications;Jan2006, Vol. 33 Issue 1, p73 

    The gradient method for the symmetric positive definite linear system Ax = b is as follows xk+1 = xk - akgk where gk = Axk - b is the residual of the system at xk and ak is the stepsize. The stepsize ak = 2/?1+?n is optimal in the sense that it minimizes the modulus ?I - aA?2, where ?1 and ?n...

  • Implementations of Affine Scaling Methods: Approximate Solutions of Systems of Linear Equations Using Preconditioned Conjugate Gradient Methods. Mehrotra, Sanjay // ORSA Journal on Computing;Spring92, Vol. 4 Issue 2, p103 

    The conjugate gradient method has been proposed for solving system of linear equations arising at each iteration of interior point methods. This paper studies several problems associated with developing such implementations. This includes development of a termination criteria, computation of an...

  • A REGULARIZED CONJUGATE GRADIENT METHOD FOR SYMMETRIC POSITIVE DEFINITE SYSTEM OF LINEAR EQUATIONS*[sup 1]). Zhong-zhi Bai; Shao-liang Zhang // Journal of Computational Mathematics;Jul2002, Vol. 20 Issue 4, p437 

    Presents a class of regularized conjugate gradient methods for solving the large sparse system of linear equations of which the coefficient matrix is an ill-conditioned symmetric positive definite matrix. Discussion on the covergence properties of these methods; Investigation on the new methods...

  • Semiempirical methods with conjugate gradient density matrix search to replace diagonalization... Daniels, Andrew D.; Millam, John M.; Scuseria, Gustavo E. // Journal of Chemical Physics;7/8/1997, Vol. 107 Issue 2, p425 

    Demonstrates how diagonalization and memory requirements are eliminated by using a conjugate gradient search for the density matrix with sparse matrix techniques. Provision of high accuracy energy calculations on molecules with thousand of atoms possible on the typical workstation; Examples on...

  • Three-dimensional homogeneous Lorentzian metrics with prescribed Ricci tensor. Calvaruso, Giovanni // Journal of Mathematical Physics;Dec2007, Vol. 48 Issue 12, p123518 

    We introduce a system of partial differential equations, whose solutions permit to determine explicitly locally homogeneous Lorentzian metrics in R3 having the prescribed admissible Ricci tensor. Solutions of this system are presented for all the different models of homogeneous Lorentzian three...

  • A New Gradient Method with an Optimal Stepsize Property. Dai, Y. H.; Yang, X. Q. // Computational Optimization & Applications;Nov2005, Vol. 32 Issue 3, p73 

    The gradient method for the symmetric positive definite linear system Ax = b is as follows xk+1 = xk - ak gk where gk = Axk - b is the residual of the system at xk and ak is the stepsize. The stepsize ak = [This symbol can not be represented into ASCI format.] is optimal in the sense that it...

  • Parallel Rayleigh Quotient Optimization with FSAI-Based Preconditioning. Bergamaschi, Luca; Martínez, Angeles; Pini, Giorgio // Journal of Applied Mathematics;2012, p1 

    The present paper describes a parallel preconditioned algorithm for the solution of partial eigenvalue problems for large sparse symmetric matrices, on parallel computers. Namely, we consider the Deflation-Accelerated Conjugate Gradient (DACG) algorithm accelerated by factorized-sparse-...

  • AN EXTENDED THREE-TERM CONJUGATE GRADIENT METHOD WITH SUFFICIENT DESCENT PROPERTY. BABAIE-KAFAKI, S.; GHANBARI, R. // Miskolc Mathematical Notes;Sep2015, Vol. 16 Issue 1, p45 

    An extension of the three-term conjugate gradient method proposed by Zhang et al. is suggested. Based on an eigenvalue analysis on the search direction matrix, it is shown that the method possesses the sufficient descent property, no matter whether the line search is exact or not as well as the...

Share

Read the Article

Courtesy of THE LIBRARY OF VIRGINIA

Sorry, but this item is not currently available from your library.

Try another library?
Sign out of this library

Other Topics