Sampling Table Configurations for the Hierarchical Poisson Dirichlet Process
Changyou Chen (NICTA)
NICTA SML SEMINARDATE: 2011-10-06
TIME: 11:00:00 - 12:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
The most popular MCMC sampling algorithm for the hierarchical PDP and hierarchical Dirichlet Process is to conduct an incremental sampling based on the Chinese restaurant metaphor, which originates from the Chinese restaurant process (CRP). In this talk, with the same metaphor, we propose a new table representation for the hierarchical PDPs by introducing an auxiliary latent variable, called table indicator, to record which customer takes responsibility for starting a new table. Based on this representation, we develop a block Gibbs sampling algorithm, which can jointly sample the data item and its table contribution. We test this out on the hierarchical Dirichlet process variant of latent Dirichlet allocation (HDP-LDA). Experiment results show that the proposed algorithm outperforms the existing algorithm in both out-of-sample perplexity and convergence speed.
