Spikes and Computation in Sensory Systems
- Speaker: Dr Simon Thorpe (CNRS)
- Host: Steve Furber
- 15th April 2009 at 14:15 in Lecture Theatre 1.4, Kilburn Building
Our visual systems can detect the presence of an animal or a human in a complex natural scene in about 100 ms. This is an astonishing feat given the hardware constraints that natural selection has had to cope with. For example, neurons rarely generate more than about 100 spikes per second, and the conduction velocity of fibres within the cortex is typically only about 1-2 m.s-1. Such constraints would seem to be fatal flaws for designers of computer hardware, and yet somehow the brain gets away with it. What is the secret ? It is often assumed that neurons transmit information in the form of a rate code, where increased activation levelS are signalled by higher firing rates. But given the temporal constraints, in seems that visual processing must be possible under conditions where each neuron may only get to fire one spike. One option is to take advantage of the fact that the most strongly activated neurons will tend to fire early, and so information can be encoded in the order in which neurons fire. In this talk, I will present some arguments for believing that the order of firing of neurons can be used as a highly efficient way to encode information and perform computations in the nervous system.