horizont implements a number of topic models. Conventions from scikit-learn are followed.

The following models are implemented using Gibbs sampling.

• Latent Dirichlet allocation (Blei et al., 2003; Pritchard et al., 2000)
• (Coming soon) Logistic normal topic model
• (Coming soon) Dynamic topic model (Blei and Lafferty, 2006)

horizont.LDA implements latent Dirichlet allocation (LDA) using Gibbs sampling. The interface follows conventions in scikit-learn.

>>> import numpy as np
>>> from horizont import LDA
>>> X = np.array([[1,1], [2, 1], [3, 1], [4, 1], [5, 8], [6, 1]])
>>> model = LDA(n_topics=2, random_state=0, n_iter=100)
>>> doc_topic = model.fit_transform(X)  # estimate of document-topic distributions
>>> model.components_  # estimate of topic-word distributions

Python 2.7 or Python 3.3+ is required. The following packages are also required:

• numpy
• scipy
• scikit-learn
• futures (Python 2.7 only)

GSL is required for random number generation inside the Pólya-Gamma random variate generator. On Debian-based sytems, GSL may be installed with the command sudo apt-get install libgsl0-dev. horizont looks for GSL headers and libraries in /usr/include and /usr/lib/ respectively.

Cython is needed if compiling from source.

• Documentation: http://pythonhosted.org/horizont
• Source code: https://github.com/ariddell/horizont/
• Issue tracker: https://github.com/ariddell/horizont/issues

horizont is licensed under Version 3.0 of the GNU General Public License. See LICENSE file for a text of the license or visit http://www.gnu.org/copyleft/gpl.html.