A new simple /spl infin/OH neuron model as a principal component analyzer

A new approach to unsupervised learning in a single-layer neural network is discussed. An algorithm for unsupervised learning based on the Hebbian learning rule is presented. A simple "almost linear" neuron model is analyzed. The adopted neuron model does not perform just simple summation of the weighted inputs but also performs integration. In addition, the adopted structure represents a dynamic neural model which contain both feedforward and feedback connections between input and output. Actually, the proposed learning algorithm could be more correctly named self-supervised rather than unsupervised. It is shown that the model neuron tends to extract the principal component from a stationary input vector sequence. The usual accepted additional decaying terms for the stabilization of original Hebb rule are avoided. The implementation of the basic Hebb scheme would not lead to an unrealistic growth of the synaptic strengths, thanks to the adopted network structure.