Modelling Neural Systems
Real nervous systems are far more complex than the connectionist view of networks of simply-connected units. Chemical and electrical synaptic activity, diffusing neurotransmitters, and both short-term and long-term plasticity are all involved. It may be better to view natural nervous systems as ever-changing chemical machines, and research into networks capable of adaptive behaviour should reflect this.
The modelling neural systems group combines two strands of theoretical models of real nervous system processes; and incorporation of such processes into neural networks used in evolutionary computation experiments. One current project models the diffusion of Nitric Oxide in real brains, and incorporates an abstract version of a diffusible neuromodulator into an artificial neural network class, the GasNet, used for robot control. For further information see the The Sussex Centre for Neuroscience or email Prof. Michael O'Shea.