Evaluating dependencies among random variables.


Keywords
feature-selection, graphical-models, information-theory, machine-learning, markov-random-field, mutual-information, python
License
MIT
Install
pip install depynd==0.4.5

Documentation

depynd Build Status Documentation Status

depynd is a Python library for evaluating dependencies among random variables from data. It supports learning statistical dependencies for one-to-one, one-to-many, and many-to-many relationships, where each one corresponds to

  • mutual information (MI) estimation,
  • feature selection, and
  • graphical model structure learning,

respectively. Specifically, depynd supports MI estimation for discrete-continuous mixtures, MI-based feature selection, and structure learning of undirected graphical models (a.k.a. Markov random fields).

Here is the documentation.

Dependencies

  • Python (>=3.5)
  • NumPy (>=1.13.0)
  • SciPy
  • scikit-learn

Installation

$ pip install depynd

How to use

See notebooks.