Signal decomposition using multiscale admixture models

Matus Telgarsky; John Lafferty

doi:10.1109/ICASSP.2007.366269

Signal decomposition using multiscale admixture models

Matus Telgarsky, John Lafferty

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Admixture models are "mixtures of mixtures" that decompose an object into multiple latent components, with the component proportions varying stochastically across objects. Recent work in machine learning has successfully developed admixture models for text, and work in population genetics has developed such models to analyze complex groups of individuals having mixed ancestry. We introduce a family of graphical admixture models for decomposing a signal into multiple components based on a wavelet representation of the signal. Two models are developed, one using a fixed segmentation of the signal, another using recursive dyadic partitioning. Variational algorithms are derived for inferring mixture proportions and estimating parameters.

Original language	English (US)
Title of host publication	2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07
Pages	II449-II452
DOIs	https://doi.org/10.1109/ICASSP.2007.366269
State	Published - 2007
Event	2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07 - Honolulu, HI, United States Duration: Apr 15 2007 → Apr 20 2007

Publication series

Name	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	2
ISSN (Print)	1520-6149

Other

Other	2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07
Country/Territory	United States
City	Honolulu, HI
Period	4/15/07 → 4/20/07

Keywords

Graphical model
Recursive dyadic partitioning
Unsupervised signal segmentation and labeling
Variational inference
Wavelets

ASJC Scopus subject areas

Software
Signal Processing
Electrical and Electronic Engineering

Access to Document

10.1109/ICASSP.2007.366269

Cite this

Telgarsky, M., & Lafferty, J. (2007). Signal decomposition using multiscale admixture models. In 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07 (pp. II449-II452). Article 4217442 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2). https://doi.org/10.1109/ICASSP.2007.366269

Signal decomposition using multiscale admixture models. / Telgarsky, Matus; Lafferty, John.
2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07. 2007. p. II449-II452 4217442 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Telgarsky, M & Lafferty, J 2007, Signal decomposition using multiscale admixture models. in 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07., 4217442, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2, pp. II449-II452, 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07, Honolulu, HI, United States, 4/15/07. https://doi.org/10.1109/ICASSP.2007.366269

@inproceedings{40f200b3b99d4137b185b39b684aef41,

title = "Signal decomposition using multiscale admixture models",

abstract = "Admixture models are {"}mixtures of mixtures{"} that decompose an object into multiple latent components, with the component proportions varying stochastically across objects. Recent work in machine learning has successfully developed admixture models for text, and work in population genetics has developed such models to analyze complex groups of individuals having mixed ancestry. We introduce a family of graphical admixture models for decomposing a signal into multiple components based on a wavelet representation of the signal. Two models are developed, one using a fixed segmentation of the signal, another using recursive dyadic partitioning. Variational algorithms are derived for inferring mixture proportions and estimating parameters.",

keywords = "Graphical model, Recursive dyadic partitioning, Unsupervised signal segmentation and labeling, Variational inference, Wavelets",

author = "Matus Telgarsky and John Lafferty",

year = "2007",

doi = "10.1109/ICASSP.2007.366269",

language = "English (US)",

isbn = "1424407281",

series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

pages = "II449--II452",

booktitle = "2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07",

note = "2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07 ; Conference date: 15-04-2007 Through 20-04-2007",

}

TY - GEN

T1 - Signal decomposition using multiscale admixture models

AU - Telgarsky, Matus

AU - Lafferty, John

PY - 2007

Y1 - 2007

N2 - Admixture models are "mixtures of mixtures" that decompose an object into multiple latent components, with the component proportions varying stochastically across objects. Recent work in machine learning has successfully developed admixture models for text, and work in population genetics has developed such models to analyze complex groups of individuals having mixed ancestry. We introduce a family of graphical admixture models for decomposing a signal into multiple components based on a wavelet representation of the signal. Two models are developed, one using a fixed segmentation of the signal, another using recursive dyadic partitioning. Variational algorithms are derived for inferring mixture proportions and estimating parameters.

AB - Admixture models are "mixtures of mixtures" that decompose an object into multiple latent components, with the component proportions varying stochastically across objects. Recent work in machine learning has successfully developed admixture models for text, and work in population genetics has developed such models to analyze complex groups of individuals having mixed ancestry. We introduce a family of graphical admixture models for decomposing a signal into multiple components based on a wavelet representation of the signal. Two models are developed, one using a fixed segmentation of the signal, another using recursive dyadic partitioning. Variational algorithms are derived for inferring mixture proportions and estimating parameters.

KW - Graphical model

KW - Recursive dyadic partitioning

KW - Unsupervised signal segmentation and labeling

KW - Variational inference

KW - Wavelets

UR - http://www.scopus.com/inward/record.url?scp=34547512940&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34547512940&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2007.366269

DO - 10.1109/ICASSP.2007.366269

M3 - Conference contribution

AN - SCOPUS:34547512940

SN - 1424407281

SN - 9781424407286

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - II449-II452

BT - 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07

T2 - 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07

Y2 - 15 April 2007 through 20 April 2007

ER -

Signal decomposition using multiscale admixture models

Abstract

Publication series

Other

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this