Preconditioning Kernel Matrices

被引:0
|
作者
Cutajar, Kurt [1 ]
Osborne, Michael A. [2 ]
Cunningham, John P. [3 ]
Filippone, Maurizio [1 ]
机构
[1] EURECOM, Dept Data Sci, Biot, France
[2] Univ Oxford, Dept Engn Sci, Oxford, England
[3] Columbia Univ, Dept Stat, New York, NY 10027 USA
关键词
INFERENCE;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The computational and storage complexity of kernel machines presents the primary barrier to their scaling to large, modern, datasets. A common way to tackle the scalability issue is to use the conjugate gradient algorithm, which relieves the constraints on both storage (the kernel matrix need not be stored) and computation (both stochastic gradients and parallelization can be used). Even so, conjugate gradient is not without its own issues: the conditioning of kernel matrices is often such that conjugate gradients will have poor convergence in practice. Preconditioning is a common approach to alleviating this issue. Here we propose preconditioned conjugate gradients for kernel machines, and develop a broad range of preconditioners particularly useful for kernel matrices. We describe a scalable approach to both solving kernel machines and learning their hyperparameters. We show this approach is exact in the limit of iterations and outperforms state-of-the-art approximations for a given computational budget.
引用
收藏
页数:10
相关论文
共 50 条
  • [21] Preconditioning for a Class of Spectral Differentiation Matrices
    Weiming Cao
    Ronald D. Haynes
    Manfred R. Trummer
    Journal of Scientific Computing, 2005, 24 : 343 - 371
  • [22] Preconditioning highly indefinite and nonsymmetric matrices
    Benzi, M
    Haws, JC
    Tuma, M
    SIAM JOURNAL ON SCIENTIFIC COMPUTING, 2000, 22 (04): : 1333 - 1353
  • [23] Stochastic preconditioning for diagonally dominant matrices
    Qian, Haifeng
    Sapatnekar, Sachin S.
    SIAM JOURNAL ON SCIENTIFIC COMPUTING, 2008, 30 (03): : 1178 - 1204
  • [24] CONCENTRATION OF KERNEL MATRICES WITH APPLICATION TO KERNEL SPECTRAL CLUSTERING
    Amini, Arash A.
    Razaee, Zahra S.
    ANNALS OF STATISTICS, 2021, 49 (01): : 531 - 556
  • [25] THE KERNEL FUNCTION IN THE GEOMETRY OF MATRICES
    MITCHELL, J
    DUKE MATHEMATICAL JOURNAL, 1952, 19 (04) : 575 - 583
  • [26] THE SPECTRUM OF KERNEL RANDOM MATRICES
    El Karoui, Noureddine
    ANNALS OF STATISTICS, 2010, 38 (01): : 1 - 50
  • [27] KERNEL AND K-KERNEL SYMMETRIC INTUITIONISTIC FUZZY MATRICES
    Punithavalli, G.
    Anandhkumar, M.
    TWMS JOURNAL OF APPLIED AND ENGINEERING MATHEMATICS, 2024, 14 (03): : 1231 - 1240
  • [28] Preconditioning of block Toeplitz matrices by sine transforms
    DiBenedetto, F
    SIAM JOURNAL ON SCIENTIFIC COMPUTING, 1997, 18 (02): : 499 - 515
  • [29] On the preconditioning of matrices with skew-symmetric splittings
    Gene H. Golub
    Denis Vanderstraeten
    Numerical Algorithms, 2000, 25 : 223 - 239
  • [30] Multilevel preconditioning for perturbed finite element matrices
    Axelsson, O
    Hakopian, YR
    Kuznetsov, YA
    IMA JOURNAL OF NUMERICAL ANALYSIS, 1997, 17 (01) : 125 - 149