Second-order Approximation of Minimum Discrimination Information in Independent Component Analysis

TL;DR

This paper introduces a second-order approximation method for ICA that improves the performance of FastICA by addressing degeneracy issues with a novel approach based on minimum discrimination information.

Contribution

It proposes a new ICA algorithm leveraging second-order approximation of MDI to enhance unmixing matrix estimation and overcome limitations of existing fixed-point methods.

Findings

Validated efficiency through experiments

Outperforms other ICA algorithms

Addresses degeneracy in FastICA

Abstract

Independent Component Analysis (ICA) is intended to recover the mutually independent sources from their linear mixtures, and F astICA is one of the most successful ICA algorithms. Although it seems reasonable to improve the performance of F astICA by introducing more nonlinear functions to the negentropy estimation, the original fixed-point method (approximate Newton method) in F astICA degenerates under this circumstance. To alleviate this problem, we propose a novel method based on the second-order approximation of minimum discrimination information (MDI). The joint maximization in our method is consisted of minimizing single weighted least squares and seeking unmixing matrix by the fixed-point method. Experimental results validate its efficiency compared with other popular ICA algorithms.