Conference Paper

Convergence of the symmetrical FastICA algorithm

Neural Networks Res. Centre, Helsinki Univ. of Technol., Espoo, Finland
DOI: 10.1109/ICONIP.2002.1202844 Conference: Neural Information Processing, 2002. ICONIP '02. Proceedings of the 9th International Conference on, Volume: 3
Source: IEEE Xplore

ABSTRACT The FastICA algorithm is one of the most popular methods to solve problems in independent component analysis (ICA) and blind source separation. It has been shown experimentally that it outperforms most of the commonly used ICA algorithms in convergence speed. A rigorous convergence analysis has been presented only for the so-called one-unit case, in which just one of the rows of the separating matrix is considered. However, in the FastICA algorithm, there is also an explicit normalization step, and it may be questioned whether the extra rotation caused by the normalization will effect the convergence speed. The purpose of this paper is to show that this is not the case and the good convergence properties of the one-unit case are also shared by the full algorithm with symmetrical normalization.

0 Followers
 · 
108 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: A kind of weighted orthogonal constrained independent component analysis (ICA) algorithms with weighted orthogonorlization for achieving this constraint is proposed recently. It has been proved in the literature that weighted orthogonal constrained ICA algorithms keep the equivariance property and have much better convergence speed, separation ability and steady state misadjustment, but the convergence is not yet analyzed in the published literature. The goal of this paper is to fill this gap. Firstly, a characterization of the stationary points corresponding to these algorithms using symmetric Minimum Distance Weighted Unitary Mapping (MDWUM) for achieving the weighted orthogonorlization is obtained. Secondly, the monotonic convergence of the weighted orthogonal constrained fixed point ICA algorithms using symmetric MDWUM for convex contrast function is proved, which is further extended to nonconvex contrast functions case by adding a weighted orthogonal constraint term onto the contrast function. Together with the boundedness of contrast function, the convergence of fixed point ICA algorithms with weighted orthogonal constraint using symmetric MDWUM is implied. Simulation experiments results show that the adaptive ICA algorithms using symmetric MDWUM are better in terms of accuracy than the ones with pre-whitening, and the fixed point ICA algorithms using symmetric MDWUM converge monotonically.
    Digital Signal Processing 05/2014; 28. DOI:10.1016/j.dsp.2014.02.005 · 1.50 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The information theoretic concept of mutual information provides a general framework to evaluate dependencies between variables. Its estimation however using B-Spline has not been used before in creating an approach for Independent Component Analysis. In this paper we present a B-Spline estimator for mutual information to find the independent components in mixed signals. Tested using electroencephalography (EEG) signals the resulting BMICA (B-Spline Mutual Information Independent Component Analysis) exhibits better performance than the standard Independent Component Analysis algorithms of FastICA, JADE, SOBI and EFICA in similar simulations. BMICA was found to be also more reliable than the "renown" FastICA. .
    04/2012; 3(2). DOI:10.5121/sipij.2012.3203
  • [Show abstract] [Hide abstract]
    ABSTRACT: Using an infinite sample, the contrast function and the FastICA algorithm are deterministic. In the practical case, we have only a finite sample. Then the contrast function and the FastICA algorithm become estimators of the deterministic case. This paper provides a unified study of the deflation FastICA algorithm assuming a finite or an infinite sample. We consider four random probability distributions based on the finite sample, and construct four FastICA estimators. We show that under mild conditions, each of these estimators are equal to a local minimizer of the contrast function with respect to the underlying random probability distribution. Making use of the existing results of M-estimators, we give a rigorous analysis of the asymptotic errors of FastICA estimators. We derive five criteria for the optimal choice of the nonlinearity function.
    IEEE Transactions on Signal Processing 04/2013; 61(8):2078-2087. DOI:10.1109/TSP.2013.2243440 · 3.20 Impact Factor