Fast Sparse Cholesky Decomposition and Inversion using Nested Dissection Matrix Reordering

被引:15
|
作者
Brandhorst, Kai [1 ]
Head-Gordon, Martin [1 ,2 ]
机构
[1] Univ Calif Berkeley, Dept Chem, Berkeley, CA 94720 USA
[2] Lawrence Berkeley Natl Lab, Div Chem Sci, Berkeley, CA USA
关键词
ELECTRONIC-STRUCTURE CALCULATIONS; PLESSET CORRELATION-ENERGY; DEGREE ORDERING ALGORITHM; DENSITY-MATRIX; NULL SPACE; DIRECT OPTIMIZATION; TENSOR FORMULATION; ELIMINATION TREES; MOLECULAR-SYSTEMS; HARTREE-FOCK;
D O I
10.1021/ct100618s
中图分类号
O64 [物理化学(理论化学)、化学物理学];
学科分类号
070304 ; 081704 ;
摘要
Here we present an efficient, yet nonlinear scaling, algorithm for the computation of Cholesky factors of sparse symmetric positive definite matrices and their inverses. The key feature of this implementation is the separation of the task into an algebraic and a numeric part. The algebraic part of the algorithm attempts to find a reordering of the rows and columns which preserves at least some degree of sparsity and afterward determines the exact nonzero structure of both the Cholesky factor and its corresponding inverse. It is based on graph theory and does not involve any kind of numerical thresholding. This preprocessing then allows for a very efficient implementation of the numerical factorization step. Furthermore this approach even allows use of highly optimized dense linear algebra kernels which leads to yet another performance boost. We will show some illustrative timings of our sparse code and compare it to the standard library implementation and a recent sparse implementation using thresholding. We conclude with some comments on how to deal with positive semidefinite matrices.
引用
收藏
页码:351 / 368
页数:18
相关论文
共 50 条
  • [31] Fast QR decomposition by using blocked Gram-Schmidt methods with Cholesky factorization
    Hosoda, Yohsuke
    Hasegawa, Takemitsu
    INFORMATION-AN INTERNATIONAL INTERDISCIPLINARY JOURNAL, 2009, 12 (06): : 1193 - 1203
  • [32] Fast Fourier transform using matrix decomposition
    Zhou, Yicong
    Cao, Weijia
    Liu, Licheng
    Agaian, Sos
    Chen, C. L. Philip
    INFORMATION SCIENCES, 2015, 291 : 172 - 183
  • [33] Fast Nonnegative Matrix Factorization Using Nested ADMM Iterations
    Lu, Wu-Sheng
    2019 IEEE PACIFIC RIM CONFERENCE ON COMMUNICATIONS, COMPUTERS AND SIGNAL PROCESSING (PACRIM), 2019,
  • [34] SPARSE REPRESENTATION OF HYPERSPECTRAL DATA USING CUR MATRIX DECOMPOSITION
    Sigurdsson, Jakob
    Ulfarsson, Magnus O.
    Sveinsson, Johannes R.
    Benediktsson, Jon Atli
    2013 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM (IGARSS), 2013, : 433 - 436
  • [35] Fast Matrix Decomposition for DOSY Using the Eigenvalue Decomposition and the Difference Approximation
    Tanaka, Yuho
    Uruma, Kazunori
    Nakao, Tomoki
    Izumi, Kenya
    Utsumi, Hiroaki
    Furukawa, Toshihiro
    TENCON 2017 - 2017 IEEE REGION 10 CONFERENCE, 2017, : 1509 - 1514
  • [36] Zhang Neurodynamics for Cholesky Decomposition of Matrix Stream Using Pseudo-Inverse with Transpose of Unknown
    Zeng, Xiu
    Yang, Min
    Guo, Jinjin
    Ling, Yihong
    Zhang, Yunong
    2021 PROCEEDINGS OF THE 40TH CHINESE CONTROL CONFERENCE (CCC), 2021, : 368 - 373
  • [37] Seismic sparse-layer reflectivity inversion using basis pursuit decomposition
    Zhang, Rui
    Castagna, John
    GEOPHYSICS, 2011, 76 (06) : R147 - R158
  • [38] Fast Co-clustering Using Matrix Decomposition
    Ling, Yun
    Ye, Chongyi
    2009 ASIA-PACIFIC CONFERENCE ON INFORMATION PROCESSING (APCIP 2009), VOL 2, PROCEEDINGS, 2009, : 201 - 204
  • [39] A fast algorithm for sparse matrix computations related to inversion (vol 242, pg 915, 2013)
    Li, S.
    Wu, W.
    Darve, E.
    JOURNAL OF COMPUTATIONAL PHYSICS, 2013, 251 : 535 - 536
  • [40] Matrix and tensor completion using tensor ring decomposition with sparse representation
    Asante-Mensah, Maame G.
    Ahmadi-Asl, Salman
    Cichocki, Andrzej
    MACHINE LEARNING-SCIENCE AND TECHNOLOGY, 2021, 2 (03):