NetCenLib

NetCenLib (Network centrality library) is a tool to compute a wide range of centrality measures for a given network. The library is designed to work with Python Networkx library.

Overview

The goal of NetCenLib is to offer a comprehensive repository for implementing a broad spectrum of centrality measures. Each year, new measures are introduced through scientific papers, often with only pseudo-code descriptions, making it difficult for researchers to evaluate and compare them with existing methods. While implementations of well-known centrality measures exist, recent innovations are frequently absent. NetCenLib strives to bridge this gap. It references the renowned CentiServer portal for well-known centrality measures and their originating papers, aiming to encompass all these measures in the future.

Code structure

All custom implementations are provided under netcenlib/algorithms package. Each centrality measure is implemented in a separate file, named after the measure itself. Correspondingly, each file contains a function, named identically to the file, which calculates the centrality measure. This function accepts a NetworkX graph as input (and other params if applicable) and returns a dictionary, mapping nodes to their centrality values. Ultimately, every custom implementation is made available through the netcenlib/algorithms package.

Implemented centrality measures:

• Algebraic
• Average Distance
• Barycenter
• Betweenness
• BottleNeck
• Centroid
• Closeness
• ClusterRank
• Communicability Betweenness
• Coreness
• Current Flow Betweenness
• Current Flow Closeness
• Decay
• Degree
• Diffusion degree
• Eigenvector
• Entropy
• Geodestic k path
• Group Betweenness Centrality
• Group Closeness
• Group Degree
• Harmonic
• Heatmap
• Katz
• Hubbell
• Laplacian
• Leverage
• Lin
• Load
• Mnc
• Pagerank
• Pdi
• Percolation
• Radiality
• Rumor
• Second Order
• Semi Local
• Subgraph
• Topological
• Trophic Levels

How to use

Library can be installed using pip:

pip install netcenlib

Code usage

Provided algorithms can be executed in the following ways:

• by invoking a specific function from netcenlib.algorithms package, which computes a given centrality measure for a given graph.

import networkx as nx
import netcenlib as ncl

# Create a graph
G = nx.karate_club_graph()

# Compute degree centrality
degree_centrality = ncl.degree_centrality(G)

# Compute betweenness centrality
betweenness_centrality = ncl.betweenness_centrality(G)

# Compute closeness centrality
closeness_centrality = ncl.closeness_centrality(G)

# Compute eigenvector centrality
eigenvector_centrality = ncl.eigenvector_centrality(G)

• invoking compute_centrality method of CentralityService class, which allows to compute centrality for a given centrality measure.

from typing import Any
import networkx as nx
from networkx import Graph

from netcenlib.centrality import compute_centrality
from netcenlib.taxonomies import Centrality

g: Graph = nx.karate_club_graph()
centrality_centroid: dict[Any, float] = compute_centrality(g, Centrality.CENTROID)

This method allows you not to directly specify centrality, making it easy to compute different centralities in a loop.

Contributing

For contributing, refer to its CONTRIBUTING.md file. We are a welcoming community... just follow the Code of Conduct.

Maintainers

Project maintainers are:

• Damian Frąszczak
• Edyta Frąszczak