The bipartiteSBM User Guide

The bipartiteSBM is a Python library of a fast community inference heuristic of the bipartite Stochastic Block Model (biSBM), using the MCMC sampler or the Kernighan-Lin algorithm. It estimates the number of communities (as well as the partition) for a bipartite network.

The bipartiteSBM helps you infer the number of communities in a bipartite network.

The bipartiteSBM utilizes the Minimum Description Length principle to determine a point estimate of the numbers of communities in the biSBM that best compress the model and data. Several test examples are included.

Supported and tested on Python>=3.6.

If you have any questions, please contact tzuchi.yen@colorado.edu.

Quick start

• Why should I use bipartiteSBM?
• Installation
• Example Kernighan-Lin inference
• Example MCMC inference
• More MCMC inference
• Explore the consistency of results
• Generate synthetic networks
• Make plots
• Run codes in parallel
• How to interpret entropy?

Dataset

• Graph input format
• Node type ordering
• Southern women network
• Malaria gene-substring network
• Ancient metabolic network

Module documentation

• biSBM package
• engines package

Frequently Asked Questions (FAQ)

• Local minima
• Choosing parameters
• Detectability limit

Additional resources

• Useful slides

Acknowledgements

The bipartiteSBM is inspired and supported by the following great humans, Daniel B. Larremore, Tiago de Paula Peixoto, Jean-Gabriel Young, Pan Zhang, and Jie Tang. Thanks Valentin Haenel who helped debug and fix the Numba code.