Generalized Independent Component Analysis as Density Estimation
We propose a new generalized ICA framework in the form of a multi-layer perceptron as a density estimator. We adopt an optimization strategy based on two criteria: a minimum reconstruction error and a minimum distance from a uniform distribution. Some simulation results are also reported to validate the proposed algorithm.
KeywordsGeneralized ICA Multilayer ICA Density estimator
Unable to display preview. Download preview PDF.
- Magnus J. R. and Neudecker H.: Matrix Differential Calculus with Applications in Statistics and Econometrics, John Wiley & Sons Ltd, (1988).Google Scholar
- Marques G. C. and Almeida L. B.: Separation of nonlinear mixtures using pattern repulsion, Proc. of Int. Worksh. on ICA and BSS, Aussois, France, (1999).Google Scholar
- Palmieri F., Mattera D., and Budillon A.: Multilayer Independent Component Analysis, Proc. of Int. Worksh. on ICA and BSS, Aussois, Francia, (1999).Google Scholar
- Palmieri F. and Budillon A.: Multi-Class Independent Component Analysis (MUCICA) for Rank-deficient Distributions. Advances in Independent Component Analysis, Ed. M. Girolami, Springer Verlag, (2000).Google Scholar
- Principe J.C., Xu D., Fisher J.: Information Theoretic Learning. Unsupervised Adaptive Filtering, vol 1, Simon Haykin Editor, Wiley, (2000).Google Scholar