Overview

PyDP is library for implementing Dirichlet Process mixture models (DPMM). The goal of PyDP is to provide a pure Python implementation of various algorithms for working DPMMs. As a design choice PyDP should have no dependencies on any libraries which are not supported by the PyPy Python interpreter.

Installation

You can install PyDP using bioconda.

conda install pydp -c bioconda

License

PyDP is licensed under the GPL v3, see the LICENSE.txt file for details.

Versions

0.2.3

• Fixed bug in mpear

0.2.2

• Added code for vector distributions

• Added code for clustering using MPEAR

0.2.1

• Fixed a bug in concentration sampler

• Fixed log_beta function to check if parameters are <= 0 and return -inf if so

0.2.0

• Changed the interface for AtomSampler to take cells instead of partitions.

• Added global parameter updating.

• Updated density interface to use caching.

• Added some new proposal functions.

0.1.5

• Fixed error in concentration sampler due to using the wrong parameterisation of the gamma prior.

0.1.4

• Fixed underflow issue in precision update for Gaussian model.

0.1.3

• Added code for Gaussian models.

• Added wrapper class for DP sampler.

0.1.2

• Added GPL license informtation.

Installation

Installation is the standard python setup.py install.

Dependencies

Required

• None

Optional

• SymPy >= 0.7.1 - Used for some of the diagnostic tools to compute the chi-square distribution.