2011/learn11/KelloBranching

Asynchronous, Event-Driven Critical Branching Networks

A Tutorial Presented By 'Chris Kello'

A wide range of scaling laws have been reported in neural and behavioral activity, including fractal spike trains, neuronal avalanches, 1/f scaling, power law distributed interspike intervals, and heavy-tailed response times. I will present a simple network of leaky integrate-and-fire neurons that simulates all these scaling laws, at multiple levels of analysis. The synapses include two binary switch variables, plus local asynchronous rules for flipping them, in order for the network to approach critical branching. These switches are responsible for simulating the data, and are shown to work in tandem with a basic mechanism of spike timing dependent plasticity. Time permitting, I will discuss the potential usefulness of critical branching for neural computation.