A sparse optimization approach to supervised NMF based on convex analytic method

Yu Morikawa, Masahiro Yukawa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)

Abstract

In this paper, we propose a novel scheme to supervised nonnegative matrix factorization (NMF). We formulate the supervised NMF as a sparse optimization problem assuming the availability of a set of basis vectors, some of which are irrelevant to a given matrix to be decomposed. The proposed scheme is presented in the context of music transcription and musical instrument recognition. In addition to the nonnegativity constraint, we introduce three regularization terms: (i) a block ℓ1 norm to select relevant basis vectors, and (ii) a temporal-continuity term plus the popular ℓ1 norm to estimate correct activation vectors. We present a state-of-the-art convex-analytic iterative solver which ensures global convergence. The number of basis vectors to be actively used is obtained as a consequence of optimization. Simulation results show the efficacy of the proposed scheme both in the case of perfect/imperfect basis matrices.

Original languageEnglish
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Pages6078-6082
Number of pages5
DOIs
Publication statusPublished - 2013 Oct 18
Externally publishedYes
Event2013 38th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2013 - Vancouver, BC, Canada
Duration: 2013 May 262013 May 31

Other

Other2013 38th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2013
CountryCanada
CityVancouver, BC
Period13/5/2613/5/31

Fingerprint

Factorization
Musical instruments
Transcription
Chemical activation
Availability

Keywords

  • convex analysis
  • sparse optimization
  • supervised nonnegative matrix factorization

ASJC Scopus subject areas

  • Signal Processing
  • Software
  • Electrical and Electronic Engineering

Cite this

Morikawa, Y., & Yukawa, M. (2013). A sparse optimization approach to supervised NMF based on convex analytic method. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (pp. 6078-6082). [6638832] https://doi.org/10.1109/ICASSP.2013.6638832

A sparse optimization approach to supervised NMF based on convex analytic method. / Morikawa, Yu; Yukawa, Masahiro.

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2013. p. 6078-6082 6638832.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Morikawa, Y & Yukawa, M 2013, A sparse optimization approach to supervised NMF based on convex analytic method. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings., 6638832, pp. 6078-6082, 2013 38th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2013, Vancouver, BC, Canada, 13/5/26. https://doi.org/10.1109/ICASSP.2013.6638832
Morikawa Y, Yukawa M. A sparse optimization approach to supervised NMF based on convex analytic method. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2013. p. 6078-6082. 6638832 https://doi.org/10.1109/ICASSP.2013.6638832
Morikawa, Yu ; Yukawa, Masahiro. / A sparse optimization approach to supervised NMF based on convex analytic method. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2013. pp. 6078-6082
@inproceedings{87f3003945404228a4e0095569c01b26,
title = "A sparse optimization approach to supervised NMF based on convex analytic method",
abstract = "In this paper, we propose a novel scheme to supervised nonnegative matrix factorization (NMF). We formulate the supervised NMF as a sparse optimization problem assuming the availability of a set of basis vectors, some of which are irrelevant to a given matrix to be decomposed. The proposed scheme is presented in the context of music transcription and musical instrument recognition. In addition to the nonnegativity constraint, we introduce three regularization terms: (i) a block ℓ1 norm to select relevant basis vectors, and (ii) a temporal-continuity term plus the popular ℓ1 norm to estimate correct activation vectors. We present a state-of-the-art convex-analytic iterative solver which ensures global convergence. The number of basis vectors to be actively used is obtained as a consequence of optimization. Simulation results show the efficacy of the proposed scheme both in the case of perfect/imperfect basis matrices.",
keywords = "convex analysis, sparse optimization, supervised nonnegative matrix factorization",
author = "Yu Morikawa and Masahiro Yukawa",
year = "2013",
month = "10",
day = "18",
doi = "10.1109/ICASSP.2013.6638832",
language = "English",
isbn = "9781479903566",
pages = "6078--6082",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

}

TY - GEN

T1 - A sparse optimization approach to supervised NMF based on convex analytic method

AU - Morikawa, Yu

AU - Yukawa, Masahiro

PY - 2013/10/18

Y1 - 2013/10/18

N2 - In this paper, we propose a novel scheme to supervised nonnegative matrix factorization (NMF). We formulate the supervised NMF as a sparse optimization problem assuming the availability of a set of basis vectors, some of which are irrelevant to a given matrix to be decomposed. The proposed scheme is presented in the context of music transcription and musical instrument recognition. In addition to the nonnegativity constraint, we introduce three regularization terms: (i) a block ℓ1 norm to select relevant basis vectors, and (ii) a temporal-continuity term plus the popular ℓ1 norm to estimate correct activation vectors. We present a state-of-the-art convex-analytic iterative solver which ensures global convergence. The number of basis vectors to be actively used is obtained as a consequence of optimization. Simulation results show the efficacy of the proposed scheme both in the case of perfect/imperfect basis matrices.

AB - In this paper, we propose a novel scheme to supervised nonnegative matrix factorization (NMF). We formulate the supervised NMF as a sparse optimization problem assuming the availability of a set of basis vectors, some of which are irrelevant to a given matrix to be decomposed. The proposed scheme is presented in the context of music transcription and musical instrument recognition. In addition to the nonnegativity constraint, we introduce three regularization terms: (i) a block ℓ1 norm to select relevant basis vectors, and (ii) a temporal-continuity term plus the popular ℓ1 norm to estimate correct activation vectors. We present a state-of-the-art convex-analytic iterative solver which ensures global convergence. The number of basis vectors to be actively used is obtained as a consequence of optimization. Simulation results show the efficacy of the proposed scheme both in the case of perfect/imperfect basis matrices.

KW - convex analysis

KW - sparse optimization

KW - supervised nonnegative matrix factorization

UR - http://www.scopus.com/inward/record.url?scp=84890477913&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84890477913&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2013.6638832

DO - 10.1109/ICASSP.2013.6638832

M3 - Conference contribution

SN - 9781479903566

SP - 6078

EP - 6082

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

ER -