Summary
International Symposium on Nonlinear Theory and its Applications
2005
Session Number:4-2-4
Session:
Number:4-2-4-1
Effects of synaptic plasticity on the neural network topology
Akira SAKUMA, Osamu ARAKI,
pp.321-324
Publication Date:2005/10/18
Online ISSN:2188-5079
Summary:
The purpose of this research is to clarify how the topological properties of a neural network change by a rule of synaptic plasticity. To evaluate the topology, we use three variables: (1) average path length (L) that denotes an average of minimum distance between two neurons, (2) clustering coefficient (γ) that denotes an average fraction of actual number of connections over total possible edges in the neighborhood of a neuron, and (3) group coefficient (G) that denotes an average number of connections between any groups of neurons, which is an extension of γ from a neighborhood to any groups of a neuron. We simulated one dimensional circulated neurons of integrate & fire models with Hebbian or STDP synaptic plasticity alternatively. When the initial topology is "regular" (with large γ, G, and L), L decreases in spite of maintained large values of γ and G as the learning proceeds. When the initial topology is "random" (with small γ, G, and L), G increases while L is kept small. These results suggest the synaptic plasticity changes the network topology so that each neuron transfers information more efficiently.