depynd 0.6.4

Evaluating dependencies among random variables.

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).