This is a small package for computing the entropy and the complexity of a given (set of) network(s). It implements the metrics that are proposed in the paper Mapping and discrimination of networks in the complexity-entropy plane. You can also find an open-access pre-print here.
The package contains a module tools that can be imported via from pycoment import tools. It contains functions for parsing input files of different formats, e.g., graphml and generating adjancency matrices from random network models.
The class Network can be imported via from pycoment import tools and allows to compute the corresponding network's (average) complexity and entropy. It can be instantiated with any symmetric adjacency matrix without self-loops provided as a sparse.csr_matrix. See the below example for details on further usage.
If your adjacency matrix does not fulfill the aforementioned requirements you can use pycoment.tools.clean_adjacency to make it symmetric, remove self-loops and remove isolated nodes.
A simple python setup.py develop should do.
Create a network adjacency matrix for a Watts-Strogatz network and compute its entropy and complexity
from pycoment import tools, Network
adjacency = tools.watts_strogatz_network(number_of_nodes=1000, number_neighbors=10, rewiring_probability=0.1)
net = Network(adjacency=adjacency)
entropy = net.entropy()
complexity = net.complexity()
# The value of complexity depends on the random reference network. Therefore it is advised to draw multiple samples and compute an average
avg_complexity, std_complexity = net.average_complexity(n_samples=100)
print("Entropy:", entropy)
print("Complexity:", complexity)
print("Average Complexity:", avg_complexity, "+/-", std_complexity)