Versatility

This package implements versatility (Shinn et al., 2017), which describes how closely affiliated a node is with a network community structure. It is written in Python3, and can only be guaranteed to work there. (This MAY work in Python2 if you import __future__ but this is untested... see code for details.)

Install with:

pip3 install versatility

Alternatively, clone the git repo and install with:

python3 setup.py install

Dependencies:

• Python3
• networkx
• Scipy (including numpy and matplotlib)
• bctpy: The module "bct" is bctpy, a port of the Brain Connectivity Toolbox to Python. The latest version supports Python3, and can be installed most easily with "pip install bcpty".

See function help for full documentation, but the most useful functions are:

• find_nodal_versatility - Compute the versatility of each node in a graph using a specific community detection algorithm.
• find_nodal_mean_versatility - Compute the versatility of each node across a spectrum of community detection algorithm parameters (most notably the resolution parameter) and find the average.
• find_optimal_gamma_curve - Find the mean and standard error of versatility across a spectrum of resolution parameters and (optionally) plot the result. This is most useful for finding the best resolution parameter, e.g. in Figure 3c of the original paper.

Here is a quick example to get you started:

import networkx
from versatility import *
G = networkx.karate_club_graph()
find_nodal_mean_versatility(G, find_communities_louvain, processors=2)
find_nodal_versatility(G, find_communities_louvain, algargs={"gamma" : 0.5})

If you use this code, please cite:

Shinn, M., Romero-Garcia, R., Seidlitz, J., Vasa, F., Vertes, P.,
Bullmore, E. (2017). Versatility of nodal affiliation to
communities. Scientific Reports 7: 4273.
doi:10.1038/s41598-017-03394-5

Copyright 2016-2019 Maxwell Shinn (maxwell.shinn@yale.edu) Available under the GNU GPLv3.