Interests in neural signalling across all scales, from ion channels to behaviour, have led to long-term collaborations with colleagues in Electronics and Computing Sciences to investigate the information processing of neuronal networks.
Biological nervous systems use many neurones operating in parallel at relatively low information processing rates, whereas conventional computers use one , or a few, processors running at high processing rates. One approach to developing more powerful computer is to move to using many processors in parallel, and a benefit of this is the production of a more powerful neuro-simulation platform. The BIMPA (Biologically-Inspired Massively Parallel Architecture) project involves collaboration with the University of Manchester to take concepts from studies of neural information processing and implement them on the new hardware. This involves caricaturing the behaviour of neurones to try to encapsulate their computation function, and also defining their connectivity. The joint aims of this project are to produce a new form of more powerful computer, and also to provide a platform on which to simulate the behaviour of neural networks containing billions of neurones. This work is timely as data sets defining extensive neuronal connectivities are beginning to be produced from a variety of imaging techniques. These ‘Connectomic’ studies need platforms like BIMPA on which we can analyse the expected information processing and predict behavioural outcomes.