Note this page will be changing as I have new ideas, or have ideas taken up by students. I would like to emphasize that I am open to any ideas or proposals - subject, of course, to them being somehow related to my own knowledge and research (so that I can be a useful to you a supervisor).
The ideas are generally open to any level (G53IDS or Masters, etc), I'll just adjust expectations to the appropriate level.
Areas of particular interest:
Note that I have approximately ordered the ideas with the 'most interesting to me' first - of course, your definition of interesting might well differ :-).
There are many well-known algorithms for solving optimisaiotn methods; but intelligently exploiting them can be tedious. Hence, in hyper-heuristics such intelligence is moved into a separate component that can be re-used. (This can be seen as an example of a good softare-engineering practice of dividing complex programs into well-understood and portable components). This is a general area; and so more than one project might well be possible in this area.
There are many important practical problems such as timetabling and scheduling for which reasonable solutions can only be found using a heuristic search process (e.g. genetic algorithms, local search, tabu search, etc). However, it is often hard to see how such algorithms work in practice. This project will design and build tools to allow insight into how such methods are working. For example, on a timetabling problems they might show a selection of timetables found so far, and highlight their differences. This would relate closely with the LANCS Initiative http://www.lancs-initiative.ac.uk/ it would form a good introduction to the area of heuristics and meta-heuristics for anyone hoping to pursue, it would also give experience in producing 'visualisation tools' for helping people gain insight into complex processes (a skill likely to be if use in many areas).
This is a general area; and so more than one project might well be possible.
Imagine you have a list of cars each one classified according to multiple criteria such as price, comfort, mile-per-gallon, etc. The project would be to develop ways to present such information graphically to a user, to help them see the patterns (more comfort means higher price?), and to help them make a choice. Try googling for multi-criteria decision making to get more idea. In business langauge this would be representing to a user the effects of tradeoffs between various desires, for the case when it is not possible to get the best of all worlds.
Academics are very concerend that people cite their papers correctly; and google scholar gives a way to get some information about this. However, names of human authors are not a good index; two people can have the smae name; one person might have differnt variations on their name. Hence, it can be difficult to do this reliably and easily. The project would be to investigate ways to help with this; e.g. maybe authors could be allocated a unique ID from some source (like the DOI system on papers), could this be incorpated in bibtex/endnote citations databases. Could the structure of results from google scholar be analysed according to co-author or key-word to break down results into clusters and so help "expplain" a set of results. Althought targetted at google scholar, the general concepts could well also be applied to other authorship issues when the underlying data is human authored; with free format, often ill-formed and with errors -- a situation likely, for example, to be found on a company intranet; hence skills learned might well be of general utility.
MORE TO COME, AS I THINK OF THEM!! You are very welcome to email me with suggestions.
Author: Andrew Parkes