Neuromorphic Systems: Past, Present and Future

When?

Wednesday 3 December 2008, 14:00 to 15:00

Where?

39BB02

Open to:

Staff, Students

Abstract: 

Neuromorphic systems are silicon/electronic implementations of neural systems. The level of these implementations varies from microscopic (for example, modelling patches of neural membrane) to modelling networks of neurons to modelling parts of functional systems (for example, elements of the visual system). In this talk, I will discuss the history of this area, talk about current research, and give some pointers to likely future directions.

Notes:

Leslie Smith did his Ph.D. in software techniques for parallel computing, but soon afterwards became much more interested in how brains (which are parallel systems, if not computers) operate. After working on neural networks in software for a number of years, he became interested in hardware for these, and collaborated with the Department of Electrical Engineering at Edinburgh University on implementing integrate-and-fire neural networks in analogue VLSI. At the same time, he was working on modelling the mammalian early auditory system in order to understand how it achieves such good results in noisy and reverberant environments (whereas the FFT-based approaches being taken by the speech recognition community did not work well in noise). Putting the hardware neural networks and the sensory processing interests together led him to work on neuromorphic systems. Although he is currently Head of Department of Computing Science and Mathematics at Stirling University, he is still (so far) maintaining his research interests.