Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Estimating Higher-Order Moments Using Symmetric Tensor Decomposition

Journal Article · · SIAM Journal on Matrix Analysis and Applications
DOI:https://doi.org/10.1137/19M1299633· OSTI ID:1725841
 [1];  [2]
  1. Univ. of Notre Dame, IN (United States)
  2. Sandia National Lab. (SNL-CA), Livermore, CA (United States)

In this paper, we consider the problem of decomposing higher-order moment tensors, i.e., the sum of symmetric outer products of data vectors. Such a decomposition can be used to estimate the means in a Gaussian mixture model and for other applications in machine learning. The dth-order empirical moment tensor of a set of p observations of n variables is a symmetric d-way tensor. Our goal is to nd a low-rank tensor approximation comprising r $$\ll$$ p symmetric outer products. The challenge is that forming the empirical moment tensor costs O(pnd) operations and O(nd) storage, which may be prohibitively expensive; additionally, the algorithm to compute the low-rank approximation costs O(nd) per iteration. Our contribution is avoiding formation of the moment tensor, computing the low-rank tensor approximation of the moment tensor implicitly using O(pnr) operations per iteration and no extra memory. This advance opens the door to more applications of higher-order moments since they can now be efficiently computed. We present numerical evidence of the computational savings and show an example of estimating the means for higher-order moments.

Research Organization:
Sandia National Laboratories (SNL-CA), Livermore, CA (United States)
Sponsoring Organization:
USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR) (SC-21); USDOE National Nuclear Security Administration (NNSA)
Grant/Contract Number:
AC04-94AL85000; NA0003525
OSTI ID:
1725841
Report Number(s):
SAND--2020-11568J; 692160
Journal Information:
SIAM Journal on Matrix Analysis and Applications, Journal Name: SIAM Journal on Matrix Analysis and Applications Journal Issue: 3 Vol. 41; ISSN 0895-4798
Publisher:
SIAMCopyright Statement
Country of Publication:
United States
Language:
English

References (19)

Numerical optimization for symmetric tensor decomposition journal April 2015
Generating Polynomials and Symmetric Tensor Decompositions journal November 2015
A comparison of algorithms for fitting the PARAFAC model journal April 2006
Anomaly detection in scientific data using joint statistical moments journal June 2019
Multivariate skewness and kurtosis measures with an application in ICA journal November 2008
Computing symmetric rank for symmetric tensors journal January 2011
Tensor decomposition in post-Hartree–Fock methods. I. Two-electron integrals and MP2 journal February 2011
Estimation and Testing of Parameters in Multivariate Laplace Distribution journal January 2005
Independent component analysis and (simultaneous) third-order tensor diagonalization journal January 2001
NPSA: Nonorthogonal Principal Skewness Analysis journal January 2020
Symmetric Tensors and Symmetric Tensor Rank journal January 2008
Efficient MATLAB Computations with Sparse and Factored Tensors journal January 2008
Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions journal January 2011
Shifted Power Method for Computing Tensor Eigenpairs journal October 2011
Efficient Computation of Higher-Order Cumulant Tensors journal January 2018
The Sample Complexity of Multireference Alignment journal January 2019
Most Tensor Problems Are NP-Hard journal November 2013
Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization journal December 1997
Cumulants journal January 2009

Similar Records

SymProp: Scaling Sparse Symmetric Tucker Decomposition via Symmetry Propagation
Conference · Sun Jun 01 00:00:00 EDT 2025 · OSTI ID:3002148

Shifted power method for computing tensor eigenvalues.
Conference · Thu Jul 01 00:00:00 EDT 2010 · OSTI ID:1021585

Shifted power method for computing tensor eigenpairs.
Technical Report · Fri Oct 01 00:00:00 EDT 2010 · OSTI ID:1005408