Community detection with spiking neural networks for neuromorphic hardware

摘要

We present results related to the performance of an algorithm for communitydetection which incorporates event-driven computation. We define a mappingwhich takes a graph G to a system of spiking neurons. Using a fully connectedspiking neuron system, with both inhibitory and excitatory synapticconnections, the firing patterns of neurons within the same community can bedistinguished from firing patterns of neurons in different communities. On arandom graph with 128 vertices and known community structure we show that byusing binary decoding and a Hamming-distance based metric, individualcommunities can be identified from spike train similarities. Using bipolardecoding and finite rate thresholding, we verify that inhibitory connectionsprevent the spread of spiking patterns.