New Threshold Updating Mechanism to Stabilize Activity of Hebbian Neuron in a Dynamic Stochastic ‘Multiple Synaptic’ Network, similar to Homeostatic Synaptic Plasticity Process

Unconstrained growth of synaptic activity and lack of references to synaptic depression in Hebb‟s postulate have diminished its value as a learning algorithm. The existing synaptic scaling mechanisms such as weight normalization, threshold updating, and spike time dependent plasticity have greatly contributed to address these issues in Hebb‟s postulate. However, all these mechanisms are based on the networks with single synaptic connection between neurons which process according to a central clock. This article presents a new threshold updating mechanism which behaves similar to the biological process called Homeostatic synaptic plasticity process and helps Hebb‟s presynaptic neuron to stabilize its activity in a dynamic stochastic multiple synaptic network. Our modeled network had neurons that (1) processed signals in different time scales, (2) having dynamic and multiple synaptic connections between neurons. And these synapses were modeled as stochastic computational units which had computational power to calculate their own signal release probability as a function of signal arrival time to these synapses, and (3) neurons regulated their own local excitation through the threshold updating mechanism. Under these significant features of the modeled network, the examined neuron exhibited the behavior similar to the presynaptic neuron of Hebb‟s postulate when its activity synchronized with the postsynaptic neuron‟s activity. And it demonstrated the behavior similar to the Stent‟s and Lisman‟s anti-Hebbian postulates when its activity asynchronized with the activity of the postsynaptic neuron. General Terms Neural Networks