A neural network for undercomplete independent component analysis.
Wei Lu, Jagath Chandana Rajapakse · 2000
Abstract: The existing independent component neural netw orks (ICNNs) in the literature need same number of output neurons as the input nodes to achieve independence among output activations. We present a tec hnique to learn the undercomplete ICNNs to produce an output with an low er dimension than the input by using joint entropy of a multidimensionalGaussian to approximate the mutual entrop y of the output. Our approach is not restricted by the squared Jacobian matrix of outputs with respect to the inputs, and gives a general rule and some criteria to extract both super- and sub-Gaussianly distributed signals and remove the Gaussian distributed noise. Simulation results with simulated signals and audio signals are provided.