Below are the proposals funded by MathFIT as a result of the MathFIT 2002 Call (text is included to give a flavour of the intended research).
Tutorials in Optimisation and Search Methodology: A Workshop at the interface of Atrificial Intelligence and Operational research
Investigators:
Professor E. Burke, Dr G. Kendall and Mr J.D.L. Silva (University of Nottingham)
"This workshop will encourage the cross-fertilisation of ideas and the development of new interaction between researchers in the disciplines of Operational research and Computer Science. One of the major aims of the event is to help set the agenda for inter-disciplinary search/optimisation research and to help the community carry out long term interdisciplinary research aimed at developing visionary approaches to a wide range of optimisation problems including many that are vital to the smooth and efficient running of industry, commerce and the service sector. The organisation of the workshop before the MISTA 2003 conference is important because it will allow us to benefit from a large congregation of reseachers and practitioners in one of the major application areas (scheduling) of optimisation and search technology. Indeed, as part of the MISTA conference, we have planned a special workshop which will consider how Operations Research, Computer Science and Artificial Intelligence can work more closely together and how ideas can migrate from one discipline to another. This workshop is being planned as a direct result of this tutorial workshop as we are able to take advantage of the unique set of experts from accross the various disciplines that we have invited and that have confirmed."
Advanced Research Mrthods in Machine Learning and Statistical Signal Processing
Investigators:
Dr J. Winkler, Professor M. Niranjan and Dr N. Lawrence (Sheffield University)
"Statistical signal processing and Bayesian analysis are closely related and there exists an extensively developed theory of statistical inference that recognises the importance of viewing statistics conditionally, that is, observed data is treated as known rather than unknown. Bayesian methods are used in numerous application areas, including sequential modelling and adaptive learning, and non-linear and mixture methods, and this emphasizes the close connection between machine learning and statistical signal processing. It is accepted that regularisation is the most effective method for training neural networks, and that this method is preferable to architecture selection, early stopping, and training with noise. It follows, therefore, that matrix algebra is important in machine learning and the Bayesian formalism of many problems, and computational relibability requires that the accuracy and stability of matrix algorithms be considered. This is a vast topic, and the focus during the whole workshop will therefore be restricted to specific problems that occur frequently in machine learning and statistical signal processing. This linear algebraic analysis of statistical problems forms a major heavy theme of the proposed workshop."
One day meeting on Cryptographic Number Theory
Investigators:
Dr S. Blackburn and Dr S. Galbraith (Royal Holloway University of London)
"Public key cryptography continues to be a fertile research area on the boundary of mathematics and computer science and the UK has a reputation for expertise in this domain. In particular, there is a wide range of topics in computational number theory which are directly relevant for public key cryptography.This workshop is designed to increase the breadth of these topics known in detail by the UK researchers. The workshop is scheduled to coincide with the (EPSRC funded) visit of Igor Shparlinski (Sydney) to Royal Holloway. Our aim is to foster new interactions between UK researchers and selected researchers in public key cryptography."