/pybn

A simple python library for Bayesian Network modelling and inference

Primary LanguagePython

pybn

(Descontinued): please, try https://github.com/pgmpy/pgmpy

A simple python library for Bayesian Network modelling and inference

Features:

  • A Directed Acyclic Graph (DAG) class with following functions: parents, children, ancestors, descendants, all v-structures, moralize.
  • An Undirected Graph implementation.
  • A a-Separation class for testing independencies.
  • i-Separation, an alternative method for testing independencies ina DAG, which considers inaugural variables and its descendants and is faster in bigger netoworks.
  • Conditional Probability Table (CPT) implementation with multiplication, division, marginalization, among other operations.
  • Elimination Ordering (Min-Neighbor, Min-Weight, Min-Filll, Weithed-Min-Fill)
  • Variable Eimination (removing barren variables, independent by evidence variables, creating one tables of new root variables, and so on).

Utilities:

  • Load network from BIF files.