Computing research seminars, 2005. All seminars to be held in Pimlott Lecture Theatre, (Ben Pimlott Building), unless otherwise stated.

For directions to Goldsmiths College see: http://www.goldsmiths.ac.uk/find-us/

To be added to the seminar mailing list, please contact Mark Bishop by email: m.bishop@gold.ac.uk

Tuesday, 8^th February, 1pm

Where do heuristics come from? Using abstraction to speed up search

Professor Rob Holte
University of Alberta
http://www.cs.ualberta.ca/~holte/

Abstract: This talk is aimed at AI researchers and students interested in methods for creating heuristics to be used by algorithms such as A* and IDA*. The talk will focus primarily on pattern databases, a general-purpose technique for defining heuristics that has revolutionized heuristic search, leading to breakthroughs such as the first optimal solutions for random instances of Rubik's Cube. Related techniques, such as hierarchical search, will also be discussed. Pattern databases store a heuristic in the form of a lookup table, with the heuristic itself being defined as the true distance from a state to the goal in an abstract (simplified) version of the state space. Using abstraction to define a heuristic has two key advantages: (1) very little domain-specific knowledge is needed to define accurate heuristics, which is particularly important for novel or one-time problems; and (2) numerous different high-quality heuristics can be created for a give state space. The talk will start with basic concepts and proceed through to the most recent advances.

Robert Holte's is Professor of Computing at the University of Alberta, Canada. His current research interests are in: machine learning; abstraction and search; Combinatorial auctions and intelligent information access. Robert is editor of the Machine Learning journal.

Monday, 21^st February, 4-5pm

3D Technologies Applied to Cultural Heritage

Guy Godin
National Research Council of Canada
http://www.iit-iti.nrc-cnrc.gc.ca/personnel/godin_guy_e.html

Abstract: 3D imaging, modeling and visualisation technologies have been an active research area of computer vision and graphics. Applications to cultural heritage have emerged as one of the key focus of interest in this field: they provide a wide spectrum of sizes, complexity, materials, and tasks for 3D techniques. The National Research Council of Canada has been actively involved in this world, and a participant in several collaborative projects. In this talk, we will describe work in development of active optical 3D sensors, of processing algorithms for shape and appearance modeling as well as for analytical tasks, and of high-resolution visualisation techniques. In parallel to our laboratory research, we performed on-site tests and demonstration projects in the area of cultural heritage as part of the validation of the technologies. The talk will provide an overview of these projects and of lessons learned, as well as dicuss research perspectives in the area.

Guy Godin, Senior Research Officer, Visual Information Technology Group, National Research Council of Canada Ottawa, Ontario, Canada.

Wednesday, 27^th April, 5-6pm

How to Destroy Manufacturing Industry

Adrian Bower
University of Bath, UK
http://staff.bath.ac.uk/ensab/replicator/

Abstract: In the mid twentieth century John von Neumann proposed a Universal Constructor - a machine that could copy itself. Since then a number of people have realised his idea, both in simulation, and physically. However, in the case of physical implementations, all current systems require a supply of very complicated and intricate building blocks. Rapid Prototyping is the engineer's name for 3D printing - the technology that allows a computer to print out three-dimensional plastic parts directly. This has been growing in industrial importance for the last 15 years, and many different ways of making such machines have been developed. Rapid prototyping machines are now at a stage where they should be able to manufacture many of their own component parts. What they cannot do, of course, is to put those parts together. But if a rapid prototyping machine were to be designed so that it could make its own parts and also were to be designed for ease of assembly, it should be well within the abilities of a technically competent person armed with some simple tools to put it together and to commission it. This talk will describe the RepRap project: REPlicating RAPid Prototyper. This project is attempting to build a rapid prototyped universal constructor and then to give all the results away free under the GNU General Public Licence.

Adrian Bower is Senior Lecturer in the Mechanical Engineering Department, University of Bath

Tuesday, 3^rd May, 4-5pm

Techniques for Computational Analysis of Triadic Harmony

Craig Sapp
CCRMA, Stanford University, USA
http://ccrma.stanford.edu/~craig/

Abstract: There are many motivations for developing automatically-generated harmonic analyses of music. First and foremost, it can save time from the tediousness of doing the analyses manually. This reason alone should make many music-theory students very happy. Secondly, it is a useful step towards giving computers an intrinsic understanding of music which is necessary for applications such as automatic performance and generative composition. For example, a straight performance of a musical score (such as a quantized MIDI file) by a computer can be quite boring and difficult to listen to. Harmonic analysis can be used to identify the underlying syntax in music which can then be used to make automatic performance more interesting. I will present a basic methodology for identifying the roots of chords in Humdrum-based musical scores. Starting with pitch-only information the computer's accuracy in identifying the root of chords is compared to human chord root identifications. Then different layers of information giving larger context to the chord are considered, such as the rhythm (metric position and duration), melodic resolution of notes, key contexts, and form.

Craig Sapp, Stanford University Centre for Computer Research in Music and Acoustics