bayes-nets

A playground for testing and comparing different inference methods on Bayes Nets.

Setup

Create conda environment with all required dependencies:

conda env create -f environment.yml

File Structure

Probability Distributions

The basic probability distribution classes (tabular and Gaussian) are located in distributions.py. These allow you to specify the parameters or tabular values for unconditional distributions.

Conditional distribution classes (tabular and Gaussian) are located in conditionals.py. For these, you must specify a mapping from evidence variables to probability distributions.

Bayes Nets

dag.py contains the implementation for a directed acyclic graph (DAG), along with useful methods like topological sort.

The BayesNet class (located in bayes_net.py) is a subclass of DAG that contains extra attributes and methods for sampling. Setting up a Bayes net involves specifying all connections and a CPT for each node, then calling build().

Full documentation on what you can do with BayesNets (sampling, fitting, inference) is here: https://paper.dropbox.com/doc/Bayes-Nets--Ay~lI2da1ow6iSvNuJsbXs3AAg-pQj20OftCqLRPLbGTW9di