Blind signal processing by the adaptive activation function neurons

The aim of this paper is to study an Information Theory based learning theo ry for neural units endowed with adaptive activation functions. The learnin g theory has the target to force the neuron to approximate the input-output transference that makes it hat (uniform) the probability density function of its output or, equivalently, that maximizes the entropy of the neuron re sponse. Then, a network of adaptive activation function neurons is studied, and the effectiveness of the new structure is tested on Independent Compon ent Analysis (ICA) problems. The new ICA neural algorithm is compared with the closely related 'Mixture of Densities' (MOD) technique by Xu et al.. Bo th simulation results and structural comparison show the new method is effe ctive and more efficient in computational complexity. (C) 2000 Elsevier Sci ence Ltd. All rights reserved.