GATSBY COMPUTATIONAL NEUROSCIENCE UNIT
UCL Logo

Charles Kemp
Department of Brain and Cognitive Sciences, MIT

 

 

Wednesday 14 September 2005

16:00

 

Seminar Room B10 (Basement)

Alexandra House, 17 Queen Square, London, WC1N 3AR

 

 

Learning Relational Systems

 

What kinds of knowledge do people use to group entities into categories?  A category can be defined in part by its place in a relational system -- a DOCTOR, for example, is someone who treats a PATIENT.  The infinite relational model (IRM) is a framework for the unsupervised discovery of relational systems.  Given data involving several sets of entities, these systems specify the kinds of entities in each set and the relations that are possible or likely between those kinds. I will apply the IRM to three problems: clustering objects and features, clustering with multiple types and relations, and learning ontologies.

(Joint work with Josh Tenenbaum and Tom Griffiths)