The Computational Intelligence research group aims
The group capitalises on the insights gained by applying new techniques based on biology and naturally occuring systems to appropriate problems in a wide range of disciplines, from understanding neural systems to control systems to signal processing to assisting dementia diagnosis. The group grew out of an earlier collaboration with Psychology (in particular the cognitive neuroscience group), but now has broadened its focus to a range of topics from time series prediction to computational neuroscience.
As can be seen from the list of projects, the group has wide ranging interests. These range from computational neuroscience to pattern recognition to signal processing, control systems and the application of neural networks, genetic algorithms and decision techniques applied to a wide range of areas.
On the computational neuroscience side, members of the group are interested in modelling neuronal microcircuits from the subsynaptic to the network level, including learning and adaption, studying low levels of the auditory system, and understanding what in vitro neurons are communicating to each other.
On the applications side, members are particularly interested in control techniques, and in applying natural (or soft) computing techniques to areas as diverse as the law, river ecology, and weather. There is work on processing time-varying signals using computational intelligence techniques ranging from higher order statistics to neural networks. Two particular application areas for this work have been audio signals and signals from neurophysiology. We are also interested in harnessing the full power of these techniques by using direct silicon implementation.
One focus is on the development of new evolutionary computation (EC) algorithms with an emphasis on applications in the bio-sciences. The novel EC techniques are typically suited to solving optimisation problems that are analytically intractable. Previous applications of this work have focussed on determining structural properties of large scale bio-molecules such as collagen, the results of which could lead to new drug developments to heal scar tissue such as burns. Recent work has focussed on developing novel evolutionary optimisation algorithms for systems where some form of directed intervention over time is required. These algorithms are most applicable to optimal control problems with the potential impact of this work therefore being very broad. For example, we have applied this approach to areas as diverse as use of bio-control agents in mushroom farming and optimal scheduling of chemotherapy treatment to maximise impact and minimise side effects. Contact: Dr David Cairns.
Another particular focus is on Complex Adaptive Systems (CAS). Such systems contain a large number of interacting entities with perhaps simple but non-linear interactions. These interactions result in global phenomena such as emergence. Certain properties of CAS make them elusive to describe and formally represent. The goal is to come up with formal methods and techniques to model and simulate CAS more effectively in various domains. To ensure that the techniques are applicable across various multi-disciplinary fields; inspiration is taken from Citation Networks, Social Network Analysis, Wireless and ad hoc Networks, Complex Sensing Environments and Biological Systems. Contact: Dr Amir Hussain.
The group has strong external links, with joint projects with the Universities of Edinburgh, Glasgow, Oxford, Strathclyde, Leicester, Newcastle, York and others in the UK, as well as the Free University, Amsterdam, Georgia Tech, USA, Australian National University, Canberra and many other international collaborations. The group has been highly active internationally and nationally. The group helps to organise the BICS series of meetings, including one held in Stirling in August/September 2004 (Biologically Inspired Cognitive Systems). It has organised a series of international workshops (EWNS1 and 2 - European Workshops on Neuromorphic Systems), and is active in organising other meetings in this area. The group has also played a major role in other meetings: ICANN (International Conference on Neural Networks), CNS (Computational NeuroScience) and NEURAP (Neural Networks and their Applications). Stirling coordinated the EPSRC network on Multiple Model based Learning Control Paradigms for Complex Systems and the EPSRC emergent computing network on Silicon and Neurobiology, which ran a number of successful workshops at the boundary between electronics and neurobiology. This area has led to more research projects (see above), and a growing interest in Neuroinformatics. Prof. Smith is a member of organising committee for the UK node of INCF: the International Neuroinformatics Coordinating Facility.
Publications are listed with other Departmental publications. Additionally, full text of these can generally be found on the principal author's web pages.