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.
- Python (>=3.5)
- NumPy (>=1.13.0)
- SciPy
- scikit-learn
$ pip install depynd
See notebooks.