Yee Whye Teh, Hal Daume III, Daniel Roy

Gatsby Computational Neuroscience Unit, UCL; University of Maryland, College Park; Massachusetts Institute of Technology

Bayesian Agglomerative Clustering with Coalescents

9:30 - 9:50am Wednesday, December 05, 2007

This is part of the Session 4: Probabilistic Models and Methods which begins at 08:30 on Wednesday December 5, 2007

W11

We introduce a new Bayesian model for hierarchical clustering based on a prior over trees called Kingman's coalescent. We develop novel greedy and sequential Monte Carlo inferences which operate in a bottom-up agglomerative fashion. We show experimentally the superiority of our algorithms over others, and demonstrate our approach in document clustering and phylolinguistics.