We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Numerical Analysis

Title: Learning Diagonal Gaussian Mixture Models and Incomplete Tensor Decompositions

Abstract: This paper studies how to learn parameters in diagonal Gaussian mixture models. The problem can be formulated as computing incomplete symmetric tensor decompositions. We use generating polynomials to compute incomplete symmetric tensor decompositions and approximations. Then the tensor approximation method is used to learn diagonal Gaussian mixture models. We also do the stability analysis. When the first and third order moments are sufficiently accurate, we show that the obtained parameters for the Gaussian mixture models are also highly accurate. Numerical experiments are also provided.
Comments: 24 pages
Subjects: Numerical Analysis (math.NA)
Cite as: arXiv:2102.04500 [math.NA]
  (or arXiv:2102.04500v2 [math.NA] for this version)

Submission history

From: Jiawang Nie [view email]
[v1] Mon, 8 Feb 2021 19:49:32 GMT (923kb,D)
[v2] Tue, 8 Jun 2021 21:58:10 GMT (924kb,D)

Link back to: arXiv, form interface, contact.