Classifier Weighted Mixture models

Elouan Argouarc'h; Francois Desbouvries; Eric Barat; Eiji Kawasaki; Thomas Dautremer

doi:10.1109/LSP.2026.3691714

Classifier Weighted Mixture models

Elouan Argouarc'h
, Francois Desbouvries
, Eric Barat
, Eiji Kawasaki
, Thomas Dautremer

Telecom Sudparis
Université Paris-Saclay

Research output: Contribution to journal › Article › peer-review

Abstract

This paper proposes an extension of standard mixture stochastic models, by replacing the constant mixture weights with functional weights defined using a classifier. Classifier Weighted Mixtures enable straightforward density evaluation, explicit sampling, and enhanced expressivity in variational estimation problems, without increasing the number of components nor the complexity of the mixture components.

Original language	English
Journal	IEEE Signal Processing Letters
DOIs	https://doi.org/10.1109/LSP.2026.3691714
Publication status	Accepted/In press - 1 Jan 2026

Keywords

classifier
mixture models
reparameterization gradients
variational estimation

Access to Document

10.1109/LSP.2026.3691714

Cite this

@article{9a56a4bf729049949d3f308adfe0ed92,

title = "Classifier Weighted Mixture models",

abstract = "This paper proposes an extension of standard mixture stochastic models, by replacing the constant mixture weights with functional weights defined using a classifier. Classifier Weighted Mixtures enable straightforward density evaluation, explicit sampling, and enhanced expressivity in variational estimation problems, without increasing the number of components nor the complexity of the mixture components.",

keywords = "classifier, mixture models, reparameterization gradients, variational estimation",

author = "Elouan Argouarc'h and Francois Desbouvries and Eric Barat and Eiji Kawasaki and Thomas Dautremer",

note = "Publisher Copyright: {\textcopyright} 1994-2012 IEEE.",

year = "2026",

month = jan,

day = "1",

doi = "10.1109/LSP.2026.3691714",

language = "English",

journal = "IEEE Signal Processing Letters",

issn = "1070-9908",

}

TY - JOUR

T1 - Classifier Weighted Mixture models

AU - Argouarc'h, Elouan

AU - Desbouvries, Francois

AU - Barat, Eric

AU - Kawasaki, Eiji

AU - Dautremer, Thomas

PY - 2026/1/1

Y1 - 2026/1/1

N2 - This paper proposes an extension of standard mixture stochastic models, by replacing the constant mixture weights with functional weights defined using a classifier. Classifier Weighted Mixtures enable straightforward density evaluation, explicit sampling, and enhanced expressivity in variational estimation problems, without increasing the number of components nor the complexity of the mixture components.

AB - This paper proposes an extension of standard mixture stochastic models, by replacing the constant mixture weights with functional weights defined using a classifier. Classifier Weighted Mixtures enable straightforward density evaluation, explicit sampling, and enhanced expressivity in variational estimation problems, without increasing the number of components nor the complexity of the mixture components.

KW - classifier

KW - mixture models

KW - reparameterization gradients

KW - variational estimation

UR - https://www.scopus.com/pages/publications/105038671137

U2 - 10.1109/LSP.2026.3691714

DO - 10.1109/LSP.2026.3691714

M3 - Article

AN - SCOPUS:105038671137

SN - 1070-9908

JO - IEEE Signal Processing Letters

JF - IEEE Signal Processing Letters

ER -

Classifier Weighted Mixture models

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this