Split-merge augmented Gibbs sampling for Hierarchical Dirichlet Processes

Santu Rana, Dinh Phung, Svetha Venkatesh

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

1 Citation (Scopus)


The Hierarchical Dirichlet Process (HDP) model is an important tool for topic analysis. Inference can be performed through a Gibbs sampler using the auxiliary variable method. We propose a splitmerge procedure to augment this method of inference, facilitating faster convergence. Whilst the incremental Gibbs sampler changes topic assignments of each word conditioned on the previous observations and model hyper-parameters, the split-merge sampler changes the topic assignments over a group of words in a single move. This allows efficient exploration of state space. We evaluate the proposed sampler on a synthetic test set and two benchmark document corpus and show that the proposed sampler enables the MCMC chain to converge faster to the desired stationary distribution.

Original languageEnglish
Title of host publicationAdvances in Knowledge Discovery and Data Mining - 17th Pacific-Asia Conference, PAKDD 2013, Proceedings
Number of pages12
EditionPART 2
Publication statusPublished - 1 Dec 2013
Externally publishedYes
EventPacific-Asia Conference on Knowledge Discovery and Data Mining 2013 - Gold Coast, Australia
Duration: 14 Apr 201317 Apr 2013
Conference number: 17th
https://link.springer.com/book/10.1007/978-3-642-37453-1 (Proceedings)

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume7819 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


ConferencePacific-Asia Conference on Knowledge Discovery and Data Mining 2013
Abbreviated titlePAKDD 2013
CityGold Coast
Internet address

Cite this