A Post-Synaptic Inhibition Recurrent Neural Network Structure and Its Application to Pattern Classification

摘要

The post-synaptic inhibition is an inhibition which is stimulated by the activity of inhibitory interneurons. When the inhibitory interneurons are stimulated by the excitatory neurons, the IPSP arised from postsynaptic membrane will inhibit the activity of post-synaptic neurons. Post-synaptic inhibition includes lateral inhibition, feedback inhibition and feedforward inhibition. In this paper, these three inhibition modalities are in deep analyzed from the angle of cognitive neuron science. A dendritic lateral inhibition Recurrent Neuron is proposed based on post-synaptic inhibition and then the Post-Synaptic Inhibition Recurrent Neural Network is constructed. Its learning algorithm is given also. By testing several benchmark classification problems, it is proved that this network structure and its learning algorithm are effective and feasible.