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

Download:

Current browse context:

eess

Change to browse by:

References & Citations

Bookmark

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

Statistics > Machine Learning

Title: Gaussian Mixture Estimation from Weighted Samples

Abstract: We consider estimating the parameters of a Gaussian mixture density with a given number of components best representing a given set of weighted samples. We adopt a density interpretation of the samples by viewing them as a discrete Dirac mixture density over a continuous domain with weighted components. Hence, Gaussian mixture fitting is viewed as density re-approximation. In order to speed up computation, an expectation-maximization method is proposed that properly considers not only the sample locations, but also the corresponding weights. It is shown that methods from literature do not treat the weights correctly, resulting in wrong estimates. This is demonstrated with simple counterexamples. The proposed method works in any number of dimensions with the same computational load as standard Gaussian mixture estimators for unweighted samples.
Comments: 7 pages, 2 (10) figures
Subjects: Machine Learning (stat.ML); Computer Vision and Pattern Recognition (cs.CV); Systems and Control (eess.SY)
MSC classes: 62G07
Cite as: arXiv:2106.05109 [stat.ML]
  (or arXiv:2106.05109v1 [stat.ML] for this version)

Submission history

From: Daniel Frisch [view email]
[v1] Wed, 9 Jun 2021 14:38:46 GMT (1704kb,D)

Link back to: arXiv, form interface, contact.