Professor Leslie S. Smith: Research Home Page

Context

Some Papers

Activities

Research statement

My research is on the general area of biologically inspired computation. This covers neuromorphic systems (particularly synthetic sensory systems), neural networks, some aspects of computational neuroscience, and some aspects of Neuroinformatics. I have gradually come from a relatively engineering-led and highly abstracted neural networks background more and more into a neurophysiologically plausible neural networks researcher: further my interest has moved into how to connect silicon and neural systems. This has led me into the realms of Neuroinformatics recently, since this areaneeds further research in order to support possible new developments in (i) really understanding how brains work in their environment (ii) being able to build bidirectional brain/computer interfaces and (iii) building machines (I don't want to call them computers) which are capable of working robustly in their environment.

Currently Funded Research Projects

• CARMEN: Code analysis, repository, and modelling for e-Neuroscience. This is a major project funded by EPSRC, including 11 UK Universities, as well as a number of commercial and non-UK groups.

  One aspect of this work is the development and testing of spike detection and sorting techniques, jointly with Dr Quian Quiroga at Leicester University. As part of this project we have a developed a set of MATLAB routines for generation of noisy spike trains.

• Auditory processing One strand of this research aims to use the fine time structure of cochlear filtered sound to develop techniques for separating out different sources of sound. At one time we spun out a company (Stirling Hearing Systems) but this company has been wound up. We have designed a new type of MEMS/CMOS microphone, and very recently have had this project funded by EPSRC. The aim of the project is to develop and test this MEMS/CMOS microphone. We have a nice lab (but no picture yet!) with a MOTU828 mark 2 and G5 Mac for sound acquisition at up to 96KSamples/second, 8 channels, and we can generate sounds at the same time. And we can adjust the reverberance of the room as well. We have worked on a comparison between some spike based techniques for ITD estimation and some cross-correlations based techniques, where both are applied at the onseet of sound in a noisy reverberant environment. In addition, I'm interested in developing ideas of spike-based features which might be invariant under reverberation, detectable in noise, and composable to enable interpretation over time.

Research Areas, currently unfunded (but not inactive)

Examination of the statistics of natural sound. This is really another strand of early auditory processing, but this time starting from the sound data itself. Such a bottom-up approach might lead to appropriate suggestions for the types of features that might be best, since they are actually present in the signal.
• Biomedical interfacing: this is an extension of earlier work (see "Talking with Nerve Cells" below). In particular, I am interested in mechanisms for more direct detection of the behaviour of excitable cells. These alter their potential using the movement of ions, so interest is in stimulation using ions, and detection of ion movement. I've coined the term Ionophysiology (instead of electrophysiology, which uses electrons) to cover this area.
• Artificial general intelligence: I am interested in what it is about neural systems that makes them different from (current) computer systems. Is there something basic about their operation that makes them different? There's a seminar (given at Edinburgh in October 2007) and I'm working up a paper for AGI-08. I did supervise a PhD (John McGeever) student in this general area some years ago: his thesis in online (postscript, PDF).
• Neural Networks: I have worked on learning rules for NNs, principal components analysis using NNs, specification techniques for NNs, and temporal neural networks, particularly spiking networks.

  There is a really basic Neural Networks introduction (really little more than HTML-ised lecture notes from a 1 hour NSYN presentation).

  In addition, there are six lectures (from 1996!) that I gave on back-propagation available.

• Recently, I have been (re-)writing a spiking neural network simulator, in Objective C, to run on MacOSX. This now has its very own web page. I am hoping to use it to work on STDP and novel ideas on data representation in spiking neural networks (such as those introduced by Izhikevich).
• Grand Challenges and The UK Common Vison Network: I have been involved with this for some time (since 2002). My particular submissions, on artificial sensing and content-based searching for digital media didn't make it to the final cut, but are, I reckon, still interesting. At least I'm still interested in them both! I am involved with two strands of the Common Vision project, namely uGC2 (Silicon meets Life) and uGC4 (Builing Brains: here's my presentation from a meeting at ISLI, Livingston, Sept26 2007)). The Common Vision website (also known as Grand Challenges in Microelectronic Design) contains further information.

Previously funded research projects

• SPIKES: Silicon Spiking Systems: Collective Parallel Computation & Adaptive Sensory Systems, a joint EPSRC project between Stirling, Glasgow and Oxford. The Research Fellow on this project was Mr. Dagmar Fraser. Although this project is finished, there is still work ongoing on early auditory pocessing, particularly spike-based early auditory processing. We have a new lab (as of spring 2007), with good sound acquisition equipment, and the capability to alter the reverberation of the lab.
• Talking with Nerve Cells: Developing communication between living vertebrate neurons and electronic systems, a joint EPSRC project between Stirling, Edinburgh and Glasgow. This project is now finished (although there is plenty more to do). The research Fellow on this project was Dr Nhamo Mtetwa who has now gone into industry. It was work on this project that led me towards my current interest in Neuroinformatics (see CARMEN, above). And also to become more interested in brain/computer inerfaces.
• Both the above grew out of the EPSRC Network: Silicon and Neurobiology: Towards a mixed Computational Paradigm. (Out-of-date website) which I co-ordinated. This network has reached the end of its journey, but has (hopefully) been influential in pushing forward research in neuron/electronic interfacing and neuromorphic systems. It resulted in two major EPSRC projects.
  
  

Research Infrastructure

• I am chair of the recently set up IEEE UKRI Computational Intelligence Chapter.
• I was one of the organisers of INCITE: Institute for Neuronal Computational Intelligence and Technology, a SHEFC funded Research Development Grant (RDG). INCITE aims to design systems which combine the speed of silicon with the robustness of human performance

Conferences

Current Conferences

I'm trying to start the organisation of a conference provisionally entitled: CI for BCI: Computational Intelligence for Brain Computer Interfacing.

We are hoping to hold Brain Inspired Cognitive Systems 2010 (BICS 10) in Madrid, Spain.

Past Conferences

In conjunction with Alan Barros (Universidade Federal do Maranhao), Igor Aleksander, Ron Chrisley and Amir Hussain, we held Brain Inspired Cognitive Systems, BICS 08, in the Brazilian town of Sao Luis, in the state of Maranhao. the proceedings of this conference are available. A book with more complete versions of some of the published papers is in preparation, for publication by Springer.

In conjunction with Igor Aleksander, Ron Chrisley and Amir Hussain, we held BICS 2006, Brain Inspired Cognitive Systems, on the Island of Lesvos, Greece at the Hotel Delfinia from October 10 - 14, 2006.

In Conjunction with Amir Hussain and Igor Aleksander, I (and NAISO) organised BICS 2004, Brain Inspired Cognitive Systems, held in Stirling University, 29 August 2004-1 September 2004. The publication CD (and other materials) from this conference is now on the web.

Alister Hamilton, Catherine Breslin and I organised the The Second European Workshop on Neural Computing (EWNS2),  held in Stirling, 3-5 September 1999. The proceedings have been published as a special isue of IJNS (volume 9 No 5).  The second workshop built on the success of the 1st European Workshop on Neuromorphic Systems, held in Stirling from 29-31 August 1997. A book from this conference Neuromorphic systems: enginering silicon from neurobiology has been published by World Scientific.