论文信息 - The Existence of Spurious Equilibrium in FastICA

The Existence of Spurious Equilibrium in FastICA

FastICA is a fast fixed-point algorithm proposed by Hyvarinen for Independent Component Analysis (ICA). For the outstanding performance such as fast convergence and robustness, it is now one of the most popular estimate methods. The existence of spurious equilibrium in FastICA is addressed in this paper, which comes from its applications in Blind Source Separation (BSS). Two issues are involved. The first is on the object function and the second is about the fixed-point algorithm, an approximate Newton’s method. Analysis shows the existence of spurious equilibrium, which is the singular point introduced during the algorithm’s derivation. Experimental results show the estimates of spurious equilibria, and improvements are proposed by revising the convergence condition.

Gang Wang | Dewen Hu

[1] E. Oja,et al. Independent Component Analysis , 2013 .

[2] Andrzej Cichocki,et al. Adaptive Blind Signal and Image Processing - Learning Algorithms and Applications , 2002 .

[3] Aapo Hyvärinen,et al. Fast and robust fixed-point algorithms for independent component analysis , 1999, IEEE Trans. Neural Networks.

[4] Aapo Hyvärinen,et al. A Fast Fixed-Point Algorithm for Independent Component Analysis , 1997, Neural Computation.

[5] Andrzej Cichocki,et al. A New Learning Algorithm for Blind Signal Separation , 1995, NIPS.

[6] Fabian J. Theis,et al. Mathematics in independent component analysis , 2003, Seventh International Symposium on Signal Processing and Its Applications, 2003. Proceedings..