pybhcd 0.1

Bayesian Hierarchical Community Discovery

bhcd: Bayesian Hierarchical Community Discovery

An efficient Bayesian nonparametric model for discovering hierarchical community structure in social networks.

Parameter Tuning

There are five parameters (alpha, beta, lambda, delta, gamma) to be tuned, lines within interval (0,1) and satisifies the following constraint.

alpha > beta lambda > delta

Build

Use CMake to build, necessary dependencies are glib and gsl. For Windows, you can use vcpkg to install the dependencies. remove lua support.

# apt install automake libtool libgsl-dev liblua5.1-dev
$ autoreconf -i
$ ./scripts/config release

Usage

Python wrapper. Set an environment variable called BHCD, pointing to the executable of bhcd. Then you can run python bhcd.py to get the hierachical tree.

$ cat >tiny.gml <<EOF
graph [
    sparse 0
    node [ id 0 label "a" ]
    node [ id 1 label "b" ]
    node [ id 2 label "c" ]
    edge [ source 0 target 1 weight 1 ]
    edge [ source 1 target 0 weight 1 ]
]
EOF
$ ./src/bhcd/bhcd -S -p tiny tiny.gml
seed: 2a23b6bb
output prefix: tiny
time: 3.010000e-04s
tree: logprob: -1.92e+00 (0:4,1:2/0:4,1:0) #intern: 2 -1.92e+00:{-1.03e+00:{a , b }, c }
$ ./scripts/bhcd_plot tiny.tree tiny.fit tiny.pdf
-4.82e+00:{-1.66e-01:{a, b}, c}

Then tiny.pdf will have a plot of the clustering.

Paper

Blundell, C., & Teh, Y. W. (2013). Bayesian hierarchical community discovery. In Advances in Neural Information Processing Systems (pp. 1601-1609).

@inproceedings{blundell2013bayesian,
  title={Bayesian hierarchical community discovery},
  author={Blundell, Charles and Teh, Yee Whye},
  booktitle={Advances in Neural Information Processing Systems},
  pages={1601--1609},
  year={2013}
}