AISTATS 2016poster63 citations

Scalable MCMC for Mixed Membership Stochastic Blockmodels

Wenzhe Li, Sungjin Ahn, Max Welling

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.

BibTeX
@InProceedings{pmlr-v51-li16d,
  title = 	 {Scalable MCMC for Mixed Membership Stochastic Blockmodels},
  author = 	 {Li, Wenzhe and Ahn, Sungjin and Welling, Max},
  booktitle = 	 {Proceedings of the 19th International Conference on Artificial Intelligence and Statistics},
  pages = 	 {723--731},
  year = 	 {2016},
  editor = 	 {Gretton, Arthur and Robert, Christian C.},
  volume = 	 {51},
  series = 	 {Proceedings of Machine Learning Research},
  address = 	 {Cadiz, Spain},
  month = 	 {09--11 May},
  publisher =    {PMLR},
  pdf = 	 {http://proceedings.mlr.press/v51/li16d.pdf},
  url = 	 {https://proceedings.mlr.press/v51/li16d.html},
  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.}
}
Scalable MCMC for Mixed Membership Stochastic Blockmodels · AISTATS 2016