TITLE

Nonnegative Matrix Factorization with Gaussian Process Priors

AUTHOR(S)
Schmidt, Mikkel N.; Laurberg, Hans
PUB. DATE
January 2008
SOURCE
Computational Intelligence & Neuroscience;2008 Supplement, Special section p1
SOURCE TYPE
Academic Journal
DOC. TYPE
Article
ABSTRACT
We present a general method for including prior knowledge in a nonnegative matrix factorization (NMF), based on Gaussian process priors. We assume that the nonnegative factors in the NMF are linked by a strictly increasing function to an underlying Gaussian process specified by its covariance function. This allows us to find NMF decompositions that agree with our prior knowledge of the distribution of the factors, such as sparseness, smoothness, and symmetries. The method is demonstrated with an example from chemical shift brain imaging.
ACCESSION #
36363880

 

Related Articles

  • Theorems on Positive Data: On the Uniqueness of NMF. Laurberg, Hans; Christensen, Mads Græsbøll; Plumbley, Mark D.; Hansen, Lars Kai; Jensen, Søren Holdt // Computational Intelligence & Neuroscience;2008 Supplement, Special section p1 

    We investigate the conditions for which nonnegative matrix factorization (NMF) is unique and introduce several theorems which can determine whether the decomposition is in fact unique or not. The theorems are illustrated by several examples showing the use of the theorems and their limitations....

  • Probabilistic Latent Variable Models as Nonnegative Factorizations. Shashanka, Madhusudana; Raj, Bhiksha; Smaragdis, Paris // Computational Intelligence & Neuroscience;2008 Supplement, Special section p1 

    This paper presents a family of probabilistic latent variable models that can be used for analysis of nonnegative data. We show that there are strong ties between nonnegative matrix factorization and this family, and provide some straightforward extensions which can help in dealing with shift...

  • Non-negative Dimensionality Reduction for Mammogram Classification. Buciu, I.; Gacsadi, A. // Journal of Electrical & Electronics Engineering;Jun2009, Vol. 2 Issue 1, p121 

    Directly classifying high dimensional data may exhibit the "curse of dimensionality'' issue that would negatively influence the classification performance with an increase in the computational load, depending also on the classifier structure. When working with classifiers not affected by this...

  • Decomposition of spontaneous movements of infants as combinations of limb synergies. Kato, Moe; Hirashima, Masaya; Oohashi, Hiroki; Watanabe, Hama; Taga, Gentaro // Experimental Brain Research;Sep2014, Vol. 232 Issue 9, p2919 

    We hypothesized that a variety of limb movements in infants, including spontaneous movements and movements during interactions with the environment, can be represented as combinations of limb synergies, which are building blocks for generating coordinated movements of multiple limbs. A...

  • POLAR DECOMPOSITION OF THE k-FOLD PRODUCT OF LEBESGUE MEASURE ON ℝn. MOGHADASI, S. REZA // Bulletin of the Australian Mathematical Society;Apr2012, Vol. 85 Issue 2, p315 

    The Blaschke–Petkantschin formula is a geometric measure decomposition of the q-fold product of Lebesgue measure on ℝn. Here we discuss another decomposition called polar decomposition by considering ℝn×⋯×ℝn as ℳn×k and using its polar...

  • A Lifting Scheme of Symmetric-antisymmetric Multiwavelet Transform for Image Coding. Chen, Jiazhong; Ju, Zengwei; Cao, Hua; Xie, Ping; Chen, Changnian; Li, Rong; Xia, Tao; Qin, Leihua // Circuits, Systems & Signal Processing;Oct2012, Vol. 31 Issue 5, p1887 

    A symmetric-antisymmetric (SA) multiwavelet lifting factorization of matrix filter banks is presented as an extension of Sweldens' traditional scalar wavelet lifting scheme. A prefilter absorbed (PA) multiwavelet lifting factorization is also presented to reduce the redundant computations in...

  • Inferential, robust non-negative matrix factorization analysis of microarray data. Paul Fogel; S. Stanley Young; Douglas M. Hawkins; Nathalie Ledirac // Bioinformatics;Jan2007, Vol. 23 Issue 1, p44 

    Motivation: Modern methods such as microarrays, proteomics and metabolomics often produce datasets where there are many more predictor variables than observations. Research in these areas is often exploratory; even so, there is interest in statistical methods that accurately point to effects...

  • Qualitative properties and classification of nonnegative solutions to -Δu = f(u) in unbounded domains when Æ’(0) < 0. Farina, Alberto; Sciunzi, Berardino // Revista Mathematica Iberoamericana;2016, Vol. 32 Issue 4, p1311 

    We consider nonnegative solutions to -Δu = ƒ(u) in unbounded Euclidean domains under zero Dirichlet boundary conditions, where ƒ is merely locally Lipschitz continuous and satisfies ƒ(0) < 0. In the half-plane, and without any other assumption on u, we prove that u is either...

  • A New Proof For DPn = CPn (n = 4). Changqing Xu; Guanghui Xu; Yuanping Zhang // Southeast Asian Bulletin of Mathematics;2006, Vol. 30 Issue 2, p377 

    This paper presents the simplest method to prove that an entrywise nonnegative, positive semidefinite matrix A of order less than 5 is completely positive so that A can be factored as A = b1b'1 + �, + bmb'm for some positive integer m, where bj's (1 = j = m) are all nonnegative vectors in...

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