New Perspectives for Relational Learning (15w5080)
Daniel Lowd (University of Oregon)
Sriraam Natarajan (Indiana University Bloomington)
David Poole (University of British Columbia)
Oliver Schulte (Simon Fraser University)
Our overall goal for the workshop is to provide a forum where researchers working with relational data can discuss the big themes for the future of the field. We want to stimulate constructive and critical discussion of the following topics.
Foster Cross-Community Collaborations. Different communities have researched relational data analysis from different perspectives. The workshop will bring together researchers from different sub-disciplines to discuss commonalities and differences in their methods. Our objective is to develop a common set of concepts, algorithms, and benchmarks. This will lay the foundation for communication and collaborations between the different research communities.
Define Challenge Problems. Researchers from different application areas will present challenge problems. These will define new tasks and set new benchmarks for relational learning. The workshop will aim for a consensus about the most promising challenge problems that should be adopted more widely. This discussion will lead to a richer evaluation methodology and make results comparable across between subfields and research groups.
Formulate New Research Questions. Participants will identify important scientific problems that go beyond the limitations of current relational analysis techniques. We invite different views from researchers for an open-ended discussion. Possibilities include using parallelism and database technologies to scale up to "big data''; representing different types of data, such as time series and real-valued variables; integrating statistical modelling and topological analysis for describing networks; and developing new languages and tools for supporting relational data analysis.
Dissemination. The Editor-in-Chief of the Machine Learning journal, who has confirmed his interest in attending, has suggested exploring the possibility of publishing a special issue on the topic of the workshop. Benefit to Canada. Many Canadian companies and organizations seek to gain actionable insights from the datasets they have collected. The workshop will further the research expertise of Canadian scientists to support digital analytics applications for large complex structured datasets. Relational learning offers a novel, challenging application area for the mathematical sciences that will provide lasting benefits for Canada.