Skip to main navigation Skip to search Skip to main content

Core consistency diagnostic aided by reconstruction error for accurate enumeration of the number of components in parafac models

  • Kefei Liu
  • , H. C. So
  • , Joao Paulo C. L. Da Costa
  • , Lei Huang

Research output: Chapters, Conference Papers, Creative and Literary WorksRGC 32 - Refereed conference paper (with host publication)peer-review

Abstract

Recently, the CORe CONsistency DIAgnostic (CORCONDIA) has attractedmore and more attention as an effective tool for determining the number of components in parallel factor analysis (PARAFAC) or Tucker 3 models. In CORCONDIA, a proper user-defined threshold is required to ensure reliable performance. The optimal threshold increases with the signal-to-noise ratio (SNR), which results in significant probability of over-enumeration of the number of components for high SNRs under fixed threshold settings. We propose to first use a threshold interval to obtain lower and upper bounds of the estimates. The estimate takes the upper bound as its initial value and is then refined based on a sequence of hypothesis tests by exploiting the reconstruction error of the PARAFAC decomposition. The proposed scheme provides accurate detection for both low and high SNRs at almost no extra computational cost. © 2013 IEEE.
Original languageEnglish
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Pages6635-6639
DOIs
Publication statusPublished - 18 Oct 2013
Event38th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2013) - Vancouver Convention Center, Vancouver, BC, Canada
Duration: 26 May 201331 May 2013

Publication series

Name
ISSN (Print)1520-6149

Conference

Conference38th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2013)
PlaceCanada
CityVancouver, BC
Period26/05/1331/05/13

Research Keywords

  • core consistency
  • multi-linear algebra
  • parallel factor analysis (PARAFAC)
  • Source enumeration

Fingerprint

Dive into the research topics of 'Core consistency diagnostic aided by reconstruction error for accurate enumeration of the number of components in parafac models'. Together they form a unique fingerprint.

Cite this