 |
 |
EPSRC Research Fellowship in Artificial Intelligence (AI) Monte Carlo Tree Search for Games and other Decision Tree Applications with Uncertainty and Hidden Information University of Bradford - School of Computing, Informatics and Media Supervisor: Professor Peter Cowling (www.scim.brad.ac.uk/~picowlin) Financial Information: Grade 8: £31,671 - £35,646 per annum. The post duration is 3 years. Start Date: 1st October 2010 (or later by agreement). Background: The desire for better Artificial Intelligence (AI) in board games, card games and, more recently, video games have been a constant driver for AI research. There is considerable current excitement over the potential of Monte Carlo Tree Search algorithms (particularly UCT (http://senseis.xmp.net/?UCT)), which have produced professional-level computer Go players in the past year, a significant step towards addressing one of the major outstanding research questions of Artificial Intelligence (http://oase.nutn.edu.tw/FUZZ_IEEE_2009/result.htm). We have received substantial funding from EPSRC to investigate the full potential of Monte Carlo Tree Search (http://gow.epsrc.ac.uk/ViewGrant.aspx?GrantRef=EP/H049061/1), in a collaboration between the University of Bradford, Imperial College London, University of Essex, University of Reykjavik (Iceland), AI Factory Ltd., Introversion Software Ltd. and Nestorgames Ltd. The Project: For the majority of problems where game/decision tree models are appropriate, particularly board and card games, real-time video games and other sequential decision problems, the tree is nondeterministic, with both hidden information and randomness. The Research Fellow will build a conceptual and a software framework for handling uncertain decision tree nodes within Monte Carlo Tree Search, particularly looking at methods which simplify and determinise random outcomes and make best use of available CPU time to produce strong AI game players/decision tools. Part of this investigation will be to consider ensemble and multi-level approaches. We will be investigating card and board games (and the fellow's preference will be important in determining the actual games to be studied) as well as related problems in General Game Playing (see http://games.stanford.edu/) and from decision support areas such as discrete event simulation. The Research Fellow: The Research Fellow will work closely with Prof Peter Cowling and two PhD students working in this area (one in post and one to be appointed), as well as maintaining regular collaborative contact with other members of the consortium. Research into Monte Carlo Tree Search is a new area, where the Fellow can make a pioneering impact and get to grips will all work to date in a relatively short time. Theoretical and experimental research skills are needed. Good programming skills are required, preferably in a range of languages such as C#/C++/Python as well as a PhD in an area related to artificial intelligence, discrete mathematics and the design of algorithms and data structures. If the PhD area, or other experience, encompasses search within trees or other combinatorial objects then this is better still. An enthusiasm for games and a knowledge of parallel programming techniques (including GPU approaches) is also highly desirable. The University, School and City: The School of Computing Informatics and Media at the University of Bradford has a strong research focus, as evidenced by our population of over 100 PhD students and the fact that 45% of our research was rated "Internationally Leading" or "World Class" by the 2008 Research Assessment Exercise (and all was recognised nationally or internationally). In 2010, all research staff and students were relocated to purpose-built facilities within the University's city-centre campus. One of the UK's largest northern cities, set amongst some of the most spectacular countryside in England and yet one of the most affordable places to live in Britain, Bradford's wonderful paradoxes create a living and working environment that is hard to beat. Bradford also has excellent transport links via Leeds/Bradford airport to London, Europe and the rest of the world, and via excellent UK road and rail connections. Academic Requirements: A good PhD in an area as above, or equivalent experience. Candidates whose first language is not English need to have IELTS 7.0 or equivalent. For informal enquiries and further information: Contact Prof Peter Cowling (email: [log in to unmask], tel.: +44 (0) 1274 234005). To Apply: See http://www.brad.ac.uk/human-resources/jobs-at-bradford/ (job ref RCI1273) for details of how to apply. All applicants are encouraged to contact Peter Cowling informally if they are considering making an application. Note the deadline (31st July 2010).