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

SEMINARS - Spring 2007

[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 46 7421, Email savi@cs.stir.ac.uk).

Talk Schedule [Top] [Abstracts]

Abstracts [Top] [Schedule]

23rd February [Schedule ]

Present and Future of Control
Dr. Leonardo Giovanini
Technical Manager & Senior Research Fellow, Industrial Control Centre, University of Strathclyde.

Abstract

Control engineers and scientists have been solving dynamical system problems based on the formulaic/systems theory approach for more than five decades. The underlying the idea in the model-based paradigm is that the entire system can be represented through one object. However, the fields of applications and computation have developed rapidly becoming complex, whilst Control itself has remained as a largely theoretical and hidden discipline.

As consequence of these changes, Control is now beginning once again to embrace the fields of artificial and distributed intelligence, pervasive and ubiquitous sensing and computing and embedded systems. This situation creates new challenges and research directions, which bring together different disciplines.

23rd February [Schedule ]

The Connected Automobile
David Marples
Telcordia and Honorary Professor, University of Stirling

Abstract

This talk will review the current state, and future, of networked automotive electronics; Starting with vehicles shipping today before going on to consider the future of the Connected Automobile and the advantages such connectivity could bring in a long term architecture where most vehicles are capable of offboard communication.

The talk is split into two parts. In the first part an overview of the current and emerging commercial landscape for connected vehicles is presented, considering some of the initiatives underway in Europe, the US and elsewhere with specific attention given to the European GST and CVIS projects and the US VII initiative. The technology, potential benefits to end users and forecast timescales for adoption are discussed. We review both the private automobile and commercial vehicle value propositions with their differing financial underpinnings. We also discuss the potential for new and novel services that are enabled by such network connectivity.

In the second part, the engineering issues associated with the development of the connected automobile are considered. Starting with the sources of data in the vehicle and moving on through the transmission, backhaul, storage and processing infrastructure needed to support vehicle applications each of the stages of the flow if data is considered in the light of the commercial initiatives described in part one. These engineering issues are cross-referenced back to the commercial initiatives considered earlier.

2nd March [Schedule ]

Efficient P2P Systems for WAN Service and Content Discovery
Alan Brown
Department of Computing Science and Mathematics, University of Stirling.

Abstract

Each day, millions of people use P2P systems such as Skype, Instant Messenger and of course content sharing systems such as Kazaa, BitTorrent, eDonkey and Kad. In a world where mobile devices operate using multiple modes of almost seamless connectivity e.g. WIFI, Bluetooth, 3G, IrDA and more, it is becoming increasingly likely that these mobile devices will want to participate in P2P networks and benefit from the rich array of services and content which will eventually be available to all. Modern content such as music, pictures and videos will be accessible along with services offering the location of a printers, petrol stations or even traffic reports. However, bandwidth for mobile nodes is expensive because mobile network operators charge a premium for the bandwidth devices use.

Because of such premiums, more efficient P2P systems are desirable for mobile nodes, where nodes can locate content and services and maintain their position in the network whilst using as little bandwidth as possible. Here I present a method which reduces network maintenance bandwidth and improves P2P overlay network performance. I also discuss the design of my new P2P Overlay system called Tork, which adjusts nodes performance according to its available bandwidth.

9th March [Schedule ]

Combining Computers and Mud - why it IS a good idea!
Martin Blunn
Department of Computing Science and Mathematics, University of Stirling.

Abstract

Archaeologists undertake a huge responsibility when starting an excavation; a responsibility to accurately record and preserve what they find in order that we may understand more about the generations before us. Often there is only a brief opportunity to observe and record an archaeological site before it is lost either to the effects of nature or, more frequently, the construction of new houses or a shopping centre.

Too frequently, the emphasis is placed on the recording of artefacts found on a site and not the soil that surrounds them. However at many sites, the soil IS the main evidence of its history. SASSA, a Soil Analysis Support System for Archaeologists, is a system being developed within the university that hopes to readdress the imbalance. By offering archaeologists a common system to record data and understand / interpret findings from a site through a soil perspective, it is hoped that more information about our past will be preserved.

The presentation will explain the background and goals of the SASSA project, discuss the software techniques used and demonstrate a prototype version of the software upon a variety of hardware platforms. (And all without getting our feet muddy!)

23rd March [Schedule ]

EDAs and Markov Networks
John McCall
Robert Gordon University .

Abstract

Estimation of Distribution Algorithms (EDA) are derived from Genetic Algorithms. In an EDA, a probabilistic model is constructed from a population of solutions. The model is then sampled repeatedly to generate a successor population of solutions, thus replacing the traditional use of genetic operators to breed children from selected solutions.

EDAs apply selective pressure by "estimating the distribution" of good solutions. A common approach is to select good solutions from the population and then construct a joint probability distribution of solution variables. Sampling the jpd then generates the next population.

Problem-dependant choices need to be made here about the jpd. Simple (univariate) EDAs assume a fully factorised jpd. More sophisticated approaches assume partial factorisations of the jpd, traditionally classified as bivariate and multivariate EDAs. A common approach is to represent the jpd using a Bayesian Network and an armoury of techniques now exists to build a Bayesian Network from a population of solutions.

In this talk, I will explore an alternative approach to modelling the distribution, which differs from the mainstream in two key aspects. Firstly the jpd is modelled as an undirected graph (Markov Network) and factorised as a Gibbs distribution. Secondly, the construction of the model does not require selection but rather uses solution fitness values directly to determine the parameters of the distribution. The approach has benefits in improving evolution but at some computational expense in building the model. I will demonstrate the costs and benefits of a Markov Network EDA by comparing with other EDAs on a range of problems.

30th March [Schedule ]

Wireless Home Networked Appliances
Madjid Merabti
Liverpool John Moores University

Abstract

In recent years, with the growth of home appliance usage and wireless communications, networked devices will become more widely used in more diverse applications than ever before. This widespread adoption will require new and novel platforms to abstract the difficulties associated with the exploitation of devices and help overcome the inherent limitations associated with bespoke development. New platforms will need to harness the interactions between devices allowing emergent behaviour to surface, whereby high-level applications are created from low-level compositions. Combining the devices we own to create new applications or provide services is challenging and there is a need to reduce the cognitive demands placed on specialists and home users alike. This will allow for better exploitation of devices and provide obvious benefits to the end user. In this talk we present a framework for managing devices and their services within home environments.

13th April [Schedule ]

Directed Intervention Crossover applied to Bio-Control Scheduling
Paul Godley
University of Stirling

Abstract

This talk describes two directed intervention crossover approaches that are applied to a bio-control dynamic system. Unlike traditional uniform crossover, both the Calculated Expanding Bin (CalEB) method and Targeted Intervention with Stochastic Selection (TInSSel) approach actively choose an intervention level and spread based on the fitness of the parents selected for crossover. Results indicate that these approaches lead to significant improvements over Uniform Crossover (UC) when a penalty is introduced for each intervention point used by the crossover algorithm.

 

20th April [Schedule ]

Feature Interactions as Inconsistencies
Luigi Logrippo
Université du Québec en Outauais, Canada

Abstract

Early research on feature interaction (FI) was based on the idea that FIs are the result of complex interleaving of features, in other words FIs were seen as process-related. The FI Contexts that were held and many research papers were based on this idea. Later it became understood that many or most FIs originate at the specification and combination stage when it is tried to combine functionalities that are logically inconsistent, and so it is useful to see FIs as logic-related. This idea applies very well in the new world of Internet Telephony, where users are able to specify the behavior of their features by means of policies, which can be seen as logical assertions on the desired behavior of a system. In this light, it has become clear that the problem of FI is a special case of the problem of consistency in the composition of software specifications, another problem that has a considerable literature. This view becomes then very general, it extends to other applications, e.g. in the area of security, and even to areas outside of computing. We will present this view, and we will show its application to call control and to the policy language APPEL. We will see that, by using the concept of pre- and post-conditions of actions, it can be determined which actions are likely to interact, which features are likely to interact, and how a logic model checker can help in this respect.

4th May [Schedule ]

Flex and Flexibility
Gregor Kiddie
INPS

Abstract

Working the the 'real world' is greatly different from a student's life. Different technologies, different processes and different attitudes. This talk will be from the perspective of a developer 'at the coal face' of an enterprise system software company involved in a multi-billion pound nationwide project. A few examples of what to expect in such an environment and how it all works (or doesn't!)

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

Last Modified: 2nd March 2007