Efficient architecture for independent component analysis

We present a new independent component analysis technique, based on the information-theoretic approach, implemented by the functional-link network, that allows to separate out mixed independent sub-Gaussian and super-Gaussian source signals. For assessing our theory we present computer simulations performed both on synthetic and real-world data, and compare the performances of the new algorithm with those of the `Mixture of Densities' approach by Xu et al.