Research : Software : SGRLD

Stochastic Gradient Riemannian Langevin Dynamics for Latent Dirichlet Allocation

Download: [tgz]

Sam Patterson and Yee Whye Teh
spatterson@gatsby.ucl.ac.uk and y.w.teh@stats.ox.ac.uk

This is free software, you can redistribute it and/or modify it under the terms of the GNU General Public License.

The GNU General Public License does not permit this software to be redistributed in proprietary programs.

This software is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

This Python code implements the stochastic gradient Riemannian Langevin dynamics (SGRLD) algorithm presented in the paper "Stochastic Gradient Riemannian Langevin Dynamics on the Probability Simplex" by Sam Patterson and Yee Whye Teh at NIPS 2013. BibTeX entry:

@inproceedings{PatTeh2013a,
  Author = {S. Patterson and Y. W. Teh},
  Booktitle = {Advances in Neural Information Processing Systems},
  Title = {Stochastic Gradient {R}iemannian {L}angevin Dynamics on the Probability Simplex},
  Year = {2013}}

We use cython for the Gibbs sampling step which means you need to compile:

$ python setup.py build_ext --inplace

The entry point is run_wiki.py which runs the algorithm on articles from wikipedia. Currently it's set to download articles as it is running which is slow. It is recommended either downloading a batch and storing for subsequent runs of the experiment, as detailed in the code, or using the xml dumps of wikipedia here:

http://dumps.wikimedia.org/enwiki/latest/

We don't have code to process those files into the required form, but there is some at

http://www.ragtag.info/2011/feb/10/processing-every-wikipedia-article/

which could be adapted to do that.