Learning Statistical Models from Relational Data
Monday, 11 August 2003
Acapulco, Mexico
This workshop will explore approaches to learning statistical models from relational data. The workshop will explore the foundations, advantages, and limitations of the surprising array of approaches that have been developed over the past decade. These include probabilistic relational models, stochastic logic programs, Bayesian logic programs, relational Bayesian networks, relational probability trees, first-order Bayesian classifiers, relational Markov models, block models and statistical relational models.
These techniques have been developed in several related, but different, subareas of artificial intelligence (reasoning under uncertainty, inductive logic programming, machine learning, and knowledge discovery and data mining) and in some areas outside of AI (e.g., databases and social network analysis). Most researchers only have exposure to one or two techniques, and no clear understanding of the relative advantages and limitations of different techniques has yet emerged. We believe this is an ideal time for a workshop that allows active researchers in this area to discuss and debate the unique challenges of learning statistical models from relational data.
This one-day workshop will consist of interactive sessions that address specific topics identified by the organizing committee (see below) rather than consisting primarily of paper presentations. Each 60-90 minute session will begin with two or three short (10-minute) presentations intended to highlight positions on a specific topic (e.g., representing probabilities or incorporating background knowledge). Prior to the workshop, participants will have access to a variety of tutorial materials provided by both organizers and participants.
Unique challenges of relational learning
Representational power of different techniques
Scalability of statistical relational model-building
Alternative methods of incorporating background knowledge
Inference and learning tasks for relational data (e.g., attribute prediction, link prediction, consolidation, entity detection, object identification and clustering)
Learning statistical models from time-changing relational data
Using statistical models to fuse relational information from noisy, heterogeneous sources
Contribution of ancillary steps to modeling (e.g., data cleaning, transformation, and querying)
Applications of relational models (e.g., social network analysis, security and law enforcement, and analysis of hypertext collections)
This workshop is intended for researchers in the areas of machine learning, knowledge discovery and data mining, information retrieval, link analysis, and social network analysis.
Participants are encouraged to submit position papers and research summaries (up to 8 pages in length) on recent and continuing research. To encourage participation but focus discussions on key topics, we also invite 2-page research synopses and position papers from participants who do not wish to submit full papers. In either case, we encourage authors to identify the discussion session under which their research/position falls.
Each submission shall be accompanied by a short statement (500 words) describing the participant's interests in the workshop topics.
Papers should be formatted according to IJCAI guidelines and should be submitted electronically in postscript, PDF, or MS Word format via e-mail.
All submissions should be sent to: srl2003@cs.umass.edu
Mar 7, 2003 Submission deadline
Mar 21, 2003 Acceptance notification
May 27, 2003 Camera-ready version of papers
Lise Getoor (co-chair)
Computer Science Department/UMIACS
AV Williams Building
University of Maryland
College Park, MD 20742
voice 301-405-2691
fax 301-405-6707
getoor@cs.umd.edu
http://www.cs.umd.edu/~getoor
David Jensen (co-chair)
Department of Computer Science
140 Governors Drive
University of Massachusetts
Amherst, MA 01003-4610
voice 413-545-9677
fax 413-545-1249
jensen@cs.umass.edu
http://www.cs.umass.edu/~jensen
James Cussens, University of York, UK
Luc De Raedt, Albert-Ludwigs-University, Germany
Pedro Domingos, University of Washington, USA
Kristian Kersting, Albert-Ludwigs-University, Germany
Stephen Muggleton, Imperial College, London, UK
Avi Pfeffer, Harvard University, USA
Taisuke Sato, Tokyo Institute of Technology, Japan
Lyle Ungar, University of Pennsylvania, USA
See the website for more information.