Scalable MCMC for Mixed Membership Stochastic Blockmodels

Open Access
Authors
Publication date 2016
Journal JMLR Workshop and Conference Proceedings
Event Conference on Artificial Intelligence and Statistics (AISTATS2016)
Volume | Issue number 51
Pages (from-to) 723-731
Organisations
  • Faculty of Science (FNWI) - Informatics Institute (IVI)
Abstract We propose a stochastic gradient Markov chain Monte Carlo (SG-MCMC) algorithm for scalable inference in mixed-membership stochastic blockmodels (MMSB). Our algorithm is based on the stochastic gradient Riemannian Langevin sampler and achieves both faster speed and higher accuracy at every iteration than the current state-of-the-art algorithm based on stochastic variational inference. In addition we develop an approximation that can handle models that entertain a very large number of communities. The experimental results show that SG-MCMC strictly dominates competing algorithms in all cases.
Document type Article
Note Artificial Intelligence and Statistics, 9-11 May 2016, Cadiz, Spain
Language English
Published at http://jmlr.org/proceedings/papers/v51/li16d.html
Downloads
li16d (Final published version)
Permalink to this page
Back