A simple biologically inspired principal component analyzer-ModH neuron model

A new approach to unsupervised learning in a single-layer neural. network is discussed. An algorithm for unsupervised learning based on Hebbian learning rule is presented. A simple neuron model is analyzed. Adopted neuron model represents dynamic neural model which contains both feed forward and feedback connections between input and output. Actually, proposed learning algorithm could be more correctly named self-supervised rather than unsupervised. The solution proposed here is a modified Hebbian rule in which the modification of the synaptic strength is proportional not to pre- and post-synaptic activity, but instead to the pre-synaptic and averaged value of post-synaptic activity. It is shown that the model neuron tends to extract the principal component from a stationary input vector sequence. Usually accepted additional decaying terms for the stabilization of original Hebb rule are avoided. Implementation of the basic Hebb scheme would not lead to unrealistic growth of the synaptic strengths, thanks to the adopted network structure.