• DocumentCode
    3326426
  • Title

    On the Unimodality of Deflation based Fast ICA Contrast

  • Author

    Gupta, Malay ; Santhanam, Balu

  • Author_Institution
    Dept. of Electr. & Comput. Eng., New Mexico Univ., Albuquerque, NM
  • fYear
    2005
  • fDate
    Oct. 28 2005-Nov. 1 2005
  • Firstpage
    237
  • Lastpage
    241
  • Abstract
    A general behavioral analysis of the contrast surface for the negentropy based ICA is performed. It is shown that while the convexity of the nonlinearity employed is certainly sufficient to guarantee unimodality of the contrast surface, a milder form of local convexity also ensures unimodality. Several classes of nonlinearities that preserve the unimodality are also introduced. Computer simulations are then used to illustrate the effect of non-convexity on the source separation capabilities of the deflation mode fast-ICA algorithm
  • Keywords
    entropy; independent component analysis; source separation; deflation mode fast-ICA algorithm; negentropy based ICA; source separation; unimodality; AWGN; Application software; Computer simulation; Convergence; Independent component analysis; Performance analysis; Principal component analysis; Robustness; Source separation; Stability;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Signals, Systems and Computers, 2005. Conference Record of the Thirty-Ninth Asilomar Conference on
  • Conference_Location
    Pacific Grove, CA
  • ISSN
    1058-6393
  • Print_ISBN
    1-4244-0131-3
  • Type

    conf

  • DOI
    10.1109/ACSSC.2005.1599740
  • Filename
    1599740