School of Computing and Mathematics
- Faculty of Natural Sciences /
- School of Computing and Mathematics /
- Research /
- Research Groups /
- Computational Intelligence and Cognitive
Computational Intelligence and Cognitive Science
We pursue a range of research that relates to Computational Intelligence and Cognitive Science. In particular our areas of interest include robotics, adaptive intelligent systems, the modelling and evaluation of human perception (hearing, speech and vision) and real-time image analysis and computer vision. In general we seek a better understanding of complex systems and the principles that underlie their emergent properties and ultimately give rise to cognition.
Some of our current and recent research activities are outlined below.
Unbounded evolution (for computational intelligence)
Alastair Channon presented the first (and so far only) closed artificial system to pass the established statistical "ALife Test" for unbounded evolutionary dynamics: an achievement identified by Bedau, Snyder and Packard as "among the very highest priorities of the field of artificial life". Earth's biosphere (through fossil-record databases) is the only other system to have passed, although many have been evaluated. This is a very significant result: we can now begin to draw generalized conclusions about open-ended evolution that were previously impossible given just the real-world example. It significantly advances our ability to generate emergent processes and structures, including complex and intelligent ones. In related work he introduced significant improvements to the test, making it well-grounded even for long-term unbounded evolution in artificial systems, through the first ever method of computing individual genes' adaptive ('normalized') evolutionary activities.
Extending the capabilities of neuro-evolutionary systems
In collaboration with his MSc students, Alastair Channon has worked on the evolution of artificial neural controllers that solve tasks requiring deliberative behaviours: tasks that cannot be solved by reactive mechanisms alone and which would traditionally have their solutions formulated in terms of search-based planning. The results of this work have demonstrated the first incremental neuro-evolutionary learning on increasingly complex versions of such tasks. Related work, with Thomas Miconi (his PhD student), was the first to demonstrate realistic co-adapted behaviours in co-evolved physically simulated articulated creatures using general purpose neurons: removing a previous barrier to the long-term evolution of new, emergent behaviours.
Theocharis Kyriacou has worked on a project called IBL (Instruction Based Learning for mobile robots). The aim of the project was to build a human-robot natural language interface. His primary role in the project was to determine and implement the primitive functions that a natural-language-able robot would need to posses at the beginning of its "life" and upon which it would build subsequent knowledge.
Theo also worked a project called RobotMODIC (Robot MODelling, IDentification and Characterisation). The project investigated various aspects of robot-environment interaction with the aim of developing a theoretical foundation in robotics. The work involved time series analysis using chaos theory, modelling of complex non-linear systems using NARMAX model structures (Nonlinear, Auto-Regressive, Moving Average models with eXogenous inputs), robot task identification and robot simulation.
Theo is currently applying numerical modelling methods in the domain of continuous seismic event localization in collaboration with the School of Physical and Geographical Sciences at Keele University.
Theocharis Kyriacou has been investigating methods of representing music using "fingerprints" that incorporate, among other, frequency and rhythm information. The aim of this investigation is to develop a method that can produce a fingerprint that would represent the musical preferences of individuals. It is hoped that this will allow a more efficient means of searching for music that an individual likes.
EPSRC funded Element-specific detection
Charles Day and Ka-Po Lam, in collaboration with Peter Haycock (iEPSAM), Tony Kearon (Criminology) and an EPSRC funded an RA (Dr. Jim Austin), have been studying how improved processing of X-ray scanner images could allow the detection of specific chemical elements. In particular we have been using neural networks to identify heavy-metals present in the target of the X-ray beam, using characteristics of their X-ray absorption signatures over a range of incident X-ray energies.
Environmental monitoring & protection
Environmental monitoring and protection (of the built and the natural environment) often involves context-sensitive, real-time, analysis of multi-dimensional, noisy and intrinsically variable data. Computational intelligence techniques are increasingly proving themselves to be a useful tool this important field and the the group is keen to promote their use further. Charles Day and John Butcher are investigating how a novel neural network technique can be used to monitor the health of reinforced concrete structures using state of the art non-destructive testing techniques.
Mining Astrophysical Datasets
Charles Day in collaboration with Pierre Maxted (iEPSAM) we secured a Nuffield Foundation Student Bursary for one of our MSci students, Andrew Dickinson, to investigate how neural networks could be used to classify data from SuperWASP observations of the night sky.
Charles Day along with David Pandyan (iLCS) and Peter Jones (iSTM), is helping to supervise Shweta Malhotra's stroke-impairment research. Shweta aims to compare the effectiveness of regression models and neural networks in predicting the impairment of upper-limb function, in patients that have recently suffered a stroke.
Colo-rectal Cancer Predictors
Colleagues from Crewe's Leighton Hospital, have provided Charles Day with some anonymised colo-rectal cancer case-data: to see if neural networks can be used to provide good predictors of treatment outcomes.
Syntactic Pattern Recognition Peter Fletcher is working on an algorithm for recognising recursively structured geometric patterns in the presence of noise, geometric distortion, and overlapping patterns. This occurs in the context of a long-term research programme into connectionist architectures for robust symbol processing. See also collaborative work on Intelligent Pattern/Biometrics Classification.
Dr Peter Fletcher
Research Assistants: Shweta Malhotra