abcd-graph 0.3.0rc3

A python library for generating ABCD graphs.

abcd-graph

A python library for generating ABCD graphs.

Installation

Using pip

pip install abcd-graph

Project available at PyPI.

From source

git clone https://github.com/AleksanderWWW/abcd-graph.git

# or - with ssh - git clone [e-mail]:AleksanderWWW/abcd-graph.git
cd abcd-graph
pip install .

Usage

from abcd_graph import ABCDGraph, ABCDParams

params = ABCDParams()
graph = ABCDGraph(params, n=1000, logger=True).build()

Parameters

• params: An instance of ABCDParams class.
• n: Number of nodes in the graph.
• logger A boolean to enable or disable logging to the console. Default is False - no logs are shown.
• callbacks: A list of instances of Callback class. Default is an empty list.

Returns

The ABCDGraph object with the generated graph.

Graph generation parameters - ABCDParams

The ABCDParams class is used to set the parameters for the graph generation.

Arguments:

Name	Type	Description	Default
gamma	float	Power-law parameter for degrees, between 2 and 3	2.5
delta	int	Min degree	5
zeta	float	Parameter for max degree, between 0 and 1	0.5
beta	float	Power-law parameter for community sizes, between 1 and 2	1.5
s	int	Min community size	20
tau	float	Parameter for max community size, between zeta and 1	0.8
xi	float	Noise parameter, between 0 and 1	0.25

Parameters are validated when the object is created. If any of the parameters are invalid, a ValueError will be raised.

Communities and edges

The ABCDGraph object has two properties that can be used to access the communities and edges of the graph.

• communities - A list of ABCDCommunity objects.
• edges - A list of tuples representing the edges of the graph.

Example:

from abcd_graph import ABCDGraph, ABCDParams

params = ABCDParams()

graph = ABCDGraph(params, n=1000, logger=True).build()

print(graph.communities)
print(graph.edges)

Communities have the following properties:

• vertices - A list of vertices in the community.
• average_degree - The average degree of the community.
• degree_sequence - The degree sequence of the community.
• empirical_xi - The empirical xi of the community.

Exporting

Exporting the graph to different formats is done via the exporter property of the Graph object.

Possible formats are:

Method	Description	Required packages	Installation command
to_networkx()	Export the graph to a networkx.Graph object.	networkx	pip install abcd[networkx]
to_igraph()	Export the graph to an igraph.Graph object.	igraph	pip install abcd[igraph]
adj_matrix	Export the graph to a numpy.ndarray object representing the adjacency matrix.
to_sparse_adjacency_matrix()	Export the graph to a scipy.sparse.csr_matrix object representing the adjacency matrix.	scipy	pip install abcd[scipy]

Example:

from abcd_graph import ABCDGraph, ABCDParams

params = ABCDParams()
graph = ABCDGraph(params, n=1000, logger=True).build()
graph_networkx = graph.exporter.to_networkx()

Callbacks

Callbacks are used to handle diagnostics and visualization of the graph generation process. They are instances of the ABCDCallback class.

Out of the box, the library provides three callbacks:

• StatsCollector - Collects statistics about the graph generation process.
• PropertyCollector - Collects properties of the graph.
• Visualizer - Visualizes the graph generation process.

Example:

from abcd_graph import ABCDGraph, ABCDParams

from abcd_graph.callbacks import StatsCollector, Visualizer, PropertyCollector

stats = StatsCollector()
vis = Visualizer()
props = PropertyCollector()
params = ABCDParams()
g = ABCDGraph(params, n=1000, logger=True, callbacks=[stats, vis, props]).build()

print(stats.statistics)

print(props.xi_matrix)

vis.draw_community_cdf()

Docker

To build a docker image containing the library, run:

docker build -t abcd-graph .

To run the image, use:

docker run -it abcd-graph /bin/bash

This will give you a terminal inside a container with the library installed.

Available are also installation commands for the additional packages:

docker build -t abcd-test --build-arg INSTALL_TYPE=igraph .

Possible values for INSTALL_TYPE are dev, matplotlib, networkx, igraph, scipy, all and extended.

Value	Packages installed
dev	pytest, pre-commit, pytest-cov
matplotlib	matplotlib
networkx	networkx
igraph	igraph
scipy	scipy
all	networkx, igraph, scipy, pytest, pre-commit, pytest-cov, matplotlib
extended	scipy, matplotlib

[!NOTE] Combinations of the above values are also possible, e.g. igraph,networkx.

[!WARNING] If you choose and option outside the available ones, the installation will still succeed, but only the base package will be installed.

Examples

The library comes with a set of examples that show how to use the library in different scenarios. You can find them in the examples directory in the format of Jupyter Notebooks.