DEPARTMENT OF COMPUTING SCIENCE AND MATHEMATICS
UNIVERSITY . COMPUTING SCIENCE . SEMINARS

SEMINARS - Autumn 2008

[Talk Schedule] [Abstracts] [Previous Seminars]

The Department of Computing Science and Mathematics presents the following seminars. Unless otherwise stated, seminars will take place in Room 4B94 of the Cottrell Building, University of Stirling from 15.00 to 16.00 on Friday afternoons during semester time. For instructions on how to get to the University, please look at the following routes.

If you would like to give a seminar to the department in future or if you need more information, please contact the seminar organiser, Savi Maharaj (Phone 01786 467431, Email savi@cs.stir.ac.uk).

Talk Schedule [Top] [Abstracts]

Abstracts [Top] [Schedule]

26th September [Schedule]

Neuromorphic Systems: Past, Present and Future
Leslie Smith
Department of Computing Science and Mathematics
University of Stirling

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 functional (for example, elements of the visual system). In this talk, I will discuss the history of this area, and give some pointers to the directions that current research is taking.

3rd October [Schedule]

Directed Intervention Crossover Approaches in Genetic Algorithms with Application to Optimal Control Problems
Paul Godley
Department of Computing Science and Mathematics
University of Stirling

Abstract

Genetic Algorithms (GAs) are a commonly used search heuristic. They are one of a range of Evolutionary Algorithms, i.e. search techniques which follow evolutionary inspired procedures to evolve solutions to problems. GAs have been successfully applied to a plethora of application areas. One area where GAs have been effective is that of deriving schedules for optimal control problems.

This talk reviews the work undertaken over the last 3 years of my PhD. into improving the performance of GAs for optimal control problems. This work has produced a number of techniques which we call "directed intervention crossover approaches", which use the fitness of parent solutions to direct the offspring production process. An overview of GAs will be followed by the results of my PhD. into applying these novel techniques to optimisation of bio-control schedules for mushroom farming and cancer chemotherapy schedules.

10th October [Schedule]

Model Checking My Favourite Novels
Muffy Calder
Department of Computing Science
University of Glasgow

Abstract

Model checking great literature - or what probabilistic reasoning using the model checker PRISM can tell us about The Curious Incident of the Dog in the Night-time and War and Peace. A must for your next book group meeting...

17th October [Schedule]

Markov Networks and Fitness Modelling in Evolutionary Algorithms [Abstract]
Alexander Brownlee
School of Computing
Robert Gordon University

Abstract

A well-known paradigm for optimisation is the genetic or evolutionary algorithm (EA). An EA maintain a population of possible solutions to a problem which converges on a global optimum using biologically-inspired selection and reproduction operators. These algorithms have been shown to perform well on a variety of hard optimisation and search problems.

A recent development in evolutionary computation is the Estimation of Distribution Algorithm (EDA) which replaces the traditional genetic reproduction operators (crossover and mutation) with the construction and sampling of a probabilistic model. While this can often represent a significant computational expense the benfit is that the model contains explicit information about the fitness function. This talk will look at one approach using a Markov network to model the distribution of fitness within the population. It will go on to explain how the resulting model can be used to reveal underlying dynamics of the problem and investigate the effect of selection on the fitness information contained within the population.

24th October [Schedule]

Abstraction and Analysis of Clinical Guidance Trees
Ken Turner
Department of Computing Science and Mathematics
University of Stirling

Abstract

This talk will present a range of enhancements to the original work of the CGT (Clinical Guidance Tree) project. This project at Stirling was undertaken by Richard Bland, Claire Beechey, Dawn Dowding and others. The project created an enhanced form of decision tree that is particularly suitable for interactive exploration by patients. A textual notation for defining guidance trees is supported by a viewer program that allows the user to interactively explore treatment options and to evaluate the likely outcomes of these.

The talk will present new work that had the following aims:

• defining an abstract notation for interactive decision trees
• formally analysing exploration errors in such trees through automated translation to Lotos (Language Of Temporal Ordering Specification)
• generating tree implementations through automated translation for an existing tree viewer
• demonstrating the approach on healthcare examples created by the CGT project.

An abstract and machine-readable notation (AD/IT - Abstract Decision/Interactive Trees) has been created for describing Clinical Guidance Trees. This is coupled with a rigorous methodology for analysing trees. Tree structure is intentionally separated from tree content so that structure and flow can be designed and evaluated before committing to detailed content for a tree. Software tools have been created to translate AD/IT tree descriptions into Lotos, XML and CGT Viewer format. These representations support formal analysis and interactive exploration of decision trees.

Through automated conversion of existing CGT trees, realistic healthcare applications have been used to validate the approach. Although healthcare has been the main focus to date, the approach is generic and of value in almost any domain where decision trees are useful.

7th November [Schedule]

Improving Associative Memory in a Network of Spiking Neurons
Russell Hunter
Department of Computing Science and Mathematics
University of Stirling

Abstract

The theory of associative memory in the brain has been investigated extensively using associative neural network models. Comparisons between neural network models of associative memory and regions of the mammalian hippocampus have been made. Biologically plausible model neural network configurations are complex systems constructed with different classes of neurons, each with a variety of properties. We compare and contrast associative memory function in a network of biologically-based spiking neurons, based on previously published results, with a simple artificial neural network. We investigate implementations of methods for improving recall under biologically realistic conditions, such as a sparsely connected network. We find that appropriately configured inhibitory circuits and active neuronal properties can compensate for the noise induced by missing connections and overlap between stored patterns.

 

14th November [Schedule]

Intracellular Modelling of Viral Infections
Perdita Stevens
School of Informatics
University of Edinburgh

Abstract

As some members of the audience will know, my experience in modelling has mostly been prescriptive modelling, e.g. UML modelling of software design. In this talk I will describe my first faltering steps into descriptive modelling; I hope that it will be of interest not only to people who already know more about biological modelling than I do, but also to anyone who has yet to take the plunge. The specific problem which I have become interested in, thanks to the enthusiasm of Dr Alain Kohl of the Laboratory for Clinical and Molecular Virology at the University of Edinburgh, is this: why does Semliki Forest Virus behave so differently in mosquitos and in vertebrates? SFV is an arbovirus, which is a contraction of "arthropod borne virus". It is carried by arthropods, in this case mosquitoes, without causing them harm as far as is known; but when the same virus infects a mammal, it causes illness. Even in the laboratory, the virus can establish a persistent low-level infection in a culture of mosquito cells, but it rapidly kills a culture of mammalian cells. Why? So far, attempts to find a convincing qualitative explanation have failed, and this may perhaps be a case where quantitative analysis is essential for understanding what's going on. We have no answers so far, but I will explain the problem and our preliminary work on it.

21st November [Schedule]

Rigorous Modelling of Web and Grid Services
Larry Tan
University of Stirling

Abstract

Grid and web services are based on the concepts of service-oriented computing. They support distribution, interoperability and composition of services in hetereogeous environments. Testing is a pragmatic and practical method to validate the functional and non-functional properties of software. Similar manual testing techniques can be applied to grid and web services.However, more automated and formally based testing of these is desirable. This talk will present a tool that allows concise specification and realisation of executable tests targeted at grid and web services.

Previous Seminar Series [Top] [Abstracts] [Schedule]

Last Modified: 25th January 2008