FABSBM

An implementation of a tractable fully Bayesian method for the stochastic block model. For more details, see [1].

[1] Kohei Hayashi, Takuya Konishi, Tatsuro Kawamoto. A Tractable Fully Bayesian Method for the Stochastic Block Model. arXiv:1602.02256

Requirements

• python 2.X
• numpy >= 1.8.2 (older version is possibly OK)
• scipy >= 0.13.3 (same here)

If you want to reproduce the figure of our experiment, following R packages are required:

• ggplot2
• scales
• Hmisc

Demo

You can run the algorithm with toy data (the true number of clusters is 4) by a command line something like this:

python run.py balanced $(N) $(K) $(seed) $(method)

• $(N) is the number of nodes of data,
• $(K) is the number of clusters of the algorithm,
• $(seed) is the random seed, and
• $(method) is the name of the algorithm.

For example, if you want to use our algorithm, try

python run.py balanced 100 4 0 FVAB

The output is formatted as follows:

balanced $(N) $(true K) $(seed) $(method) $(selected K) $(runtime) $(training log-likelihood) $(testing log-likelihood)

Reproducing experiment

You can also reproduce the toy experiment of our paper. For this, try

make -j all
make plot

Note: this can be time consuming (~1day).