The Fixed-Point Algorithm and Maximum Likelihood Estimation forIndependent Component Analysis

  • Authors:
  • Aapo Hyvärinen

  • Affiliations:
  • Helsinki University of Technology, Laboratory of Computer and Information Science, PO Box 5400, FIN-02015 HUT, Finland, e-mail: aapo.hyvarinen@hut.fi

  • Venue:
  • Neural Processing Letters
  • Year:
  • 1999

Quantified Score

Hi-index 0.01

Visualization

Abstract

The author previously introduced a fast fixed-point algorithm for independentcomponent analysis. The algorithm was derived from objective functionsmotivated by projection pursuit. In this paper, it is shown that thealgorithm is closely connected to maximum likelihood estimation aswell. The basic fixed-point algorithm maximizes the likelihood underthe constraint of decorrelation, if the score function is used as thenonlinearity. Modifications of the algorithm maximize the likelihoodwithout constraints.