i++ School Newsletter

Week commencing 20 April 2009

Previous Newsletters

Paper wins award at EuroGP 2009 conference

Sara Silva and Stephen Dignum, CSEE Research Officer, recently won the Best Paper award at the EuroGP 2009 Conference. Details of the paper are below;

Sara Silva and Stephen Dignum, Extending Operator Equalisation: Fitness Based Self Adaptive Length Distribution for Bloat Free GP, 12th European Conference on Genetic Programming, April 15-17, 2009, TÜBINGEN, Germany

Abstract - Operator equalisation is a recent bloat control technique that allows accurate control of the program length distribution during a GP run. By filtering which individuals are allowed in the population, it can easily bias the search towards smaller or larger programs. This technique achieved promising results with different predetermined target length distributions, using a conservative program length limit. Here we improve operator equalisation by giving it the ability to automatically determine and follow the ideal length distribution for each stage of the run, unconstrained by a fixed maximum limit. Results show that in most cases the new technique performs a more efficient search and effectively reduces bloat, by achieving better fitness and/or using smaller programs. The dynamics of the self adaptive length distributions are briefly analysed, and the overhead involved in following the target distribution is discussed, advancing simple ideas for improving the efficiency of this new technique.

CSEE Undergraduate scoops First Prize at BCS Women undergraduate lovelace colloquium

Adriana Alexandru, first-year BEng Computer Networks, was awarded first prize for her entry into the "Open Choice Contest", part of the "The BCS Women Undergraduate Lovelace Colloquium" conference that took place at the University of Leeds on 17 April 2009.
 
Adriana created a poster entitled “The future of Internet: Changing the Internet Protocol with the Publish-Subscribe Internet Routing Paradigm”, focusing on the PSIRP research project, which is backed by the EU and BT. In her presentation Adriana outlined the benefits of using PSIRP instead of IP and how it could improve the internet on all levels. Adriana also responded to questions from the jury on how online marketing would be affected if PSIRP was used, and how Google marketing tools (Google ads) would be affected.
 
The competition, which was judged by Alan Pollard, the president of BCS for 2008-2009.

csee student HAS ARTICLE PUBLISHED IN ieee publication

Adam Hill, an electrical engineering PhD student, has had an article on his experiences with audio engineering published in the March edition of GoldRush, the IEEE Gold Newsletter. Click here to read the article, on page 10 of the newsletter.

papers aCCEPTED

Amnon H. Eden, James H. Moor, Human Enhancement and the Intelligence Explosion, 7th European Computing And Philosophy conf.—ECAP'09 (2–4 Jul. 2009), Barcelona, Spain.

Abstract - We are concerned with the intelligence explosion as it applies to the possibility of making humans more intelligent either biologically or cybernetically, a transformation that may take place at a very rapid or ‘explosive’ pace. We focus on two issues arising from the very nature of these forecasts. The first problem concerns the difficulty to foresee what lies beyond of such a process, arising from the speed and magnitude of these changes. Can our unenhanced intelligence offer any coherent account for a radically superior intelligence? Let’s call this the Event Horizon Problem. And to the extent we can foresee the consequences of changing our intelligence, on what set of values would we make such decisions given that our human nature that grounds our values would become radically different? Let’s call this the Value Paradox.

B. Awwad Shiekh Hasan and J.Q. Gan, Sequential EM for unsupervised adaptive Gaussian mixture model based classifier, Int. Conf. on Machine Learning and Data Mining, Leipzig, Germany, 2009

Abstract - In this paper we present a sequential expectation maximization algorithm to adapt in an unsupervised manner a Gaussian mixture model for a classification problem. The goal is to adapt the Gaussian mixture model to cope with the non-stationarity in the data to classify and hence preserve the classification accuracy. Experimental results on synthetic data show that this method is able to learn the time-varying statistical features in data by adapting a Gaussian mixture model online. In order to control the adaptation method and to ensure the stability of the adapted model, we introduce an index to detect when the adaptation would fail.

Seminar Today

24 April 2009, 3.00 - 4.00pm, Room 1N1.4.1

Speaker: Dr Ashutosh Tiwari, Cranfield University (hosted by Qinfu Zhang)

Process Optimisation Using an Evolutionary Multi-objective Framework

Abstract – As industry faces the current global economic slowdown, it is looking for innovative solutions to improve its processes. Processes are traditionally improved within industry by following certain guiding principles, such as lean initiatives, business process re-engineering, total quality management and six sigma. However, most of these initiatives are manual, expensive and time-consuming due to the complex subjective nature of process re-design. This talk investigates the application of evolutionary computing for providing computer-assisted continued/automated optimisation and adaptation of processes. Process optimisation problems are complex and are characterised by their multiple stages and hierarchical nature. This talk will consist of two parts: the first part will focus on the optimisation of multi-stage production processes and the second on the optimisation of complex, non-sequential business processes. Production processes deal with product manufacture, while business processes (such as customer order taking, servicing and information exchange) support the production processes and the services offered throughout the lifecycle of a specific product. This talk will present new modelling and optimisation techniques for production and business processes. The proposed techniques identify a variety of near-optimal solutions from which one could be finally chosen based on the designer's preferences. It is also shown that the obtained solutions dominate the previous results reported in literature.

Forthcoming Seminars

1 May 2009, 3.00 - 4.00pm, Room 1N1.4.1

Speaker: Dr Gareth Howells, University of Kent (hosted by Klaus McDonald-Maier)

Adapting Weightless Neural Systems for Employment within Problem Domains Exhibiting a Large Number of Independent Pattern Classes

Abstract - Weightless RAM-based neural networks possess several significant advantages including fast learning algorithms, a simple Boolean logic based structure and efficient hardware implementations. However, such networks possess limited ability to address the scale problem exhibited by a large number of distinct pattern classes and limited training data. This talk introduces a Multi-Classifier configuration for weightless networks which addresses this issue and allows such networks to be efficiently employed within a range of problem domains.

8 May 2009, 3.00 - 4.00pm, Room 1N1.4.1

Speaker: Dr Piotr Cofta, BT (hosted by Maria Fasli)

Title - Incorporating Social Trust into Design Practices

Abstract - Social trust is an important enabler and facilitator of several applications, yet it is too often handled in an ad-hoc manner, leading to expensive failures or continuous upgrades. In order to improve the operation and adoption of systems, trust should be incorporated into design practices. The presentation analyses this challenge to demonstrate how the proposed methodology of 'designing for trust' can be used to address some of existing deficiencies of the design process. The methodology incorporates social-technical analysis with a novel approach to modelling trust networks and can detect as well as alleviate potential design defects.