cshvienna / networkinequalities Goto Github PK

Repository of the netin package to generate random scale-free graph generators proposed by members of the Network Inequality CSH group.

Home Page: https://cshvienna.github.io/NetworkInequalities/

License: Other

Python 37.51% CSS 1.65% HTML 60.84%

graph-generator homophily networks random-graphs scale-free-networks triadic-closure

networkinequalities's Introduction

NetIn

NetIn is a python package for network inference. It is based on the NetworkX package and provides a set of methods to study network inequalities. The package is currently under development and will be updated regularly.

https://static.pepy.tech/personalized-badge/netin?period=total&units=international_system&left_color=black&right_color=orange&left_text=Downloads

Website: https://www.networkinequality.com
Documentation: https://cshvienna.github.io/NetworkInequalities
Source: https://github.com/CSHVienna/NetworkInequalities
Bug reports: https://github.com/CSHVienna/NetworkInequalities/issues
GitHub Discussions: https://github.com/CSHVienna/NetworkInequalities/discussions
Mailing list: https://groups.google.com/forum/#!forum/netin-dev

Simple examples

Create an undirected network with preferential attachment and homophily.

>>> from netin import PAH
>>> G = PAH(n=200, k=2, f_m=0.2, h_MM=0.1, h_mm=0.9, seed=42)
>>> G.generate()
>>> G.info()

Create a directed network with preferential attachment and homophily.

>>> from netin import DPAH
>>> G = DPAH(n=200, f_m=0.2, d=0.02, h_MM=0.1, h_mm=0.6, plo_M=2.0, plo_m=2.0, seed=42)
>>> G.generate()
>>> G.info()

Install

Install the latest version of NetIn:

$ pip install netin

Install from source:

$ git clone
$ cd NetworkInequalities
$ pip install -e .

Bugs

Please report any bugs that you find here. Or, even better, fork the repository on GitHub and create a pull request (PR). We welcome all changes, big or small, and we will help you make the PR if you are new to git.

License

Released under Creative Commons by-nc-sa 4.0 (see LICENSE):

Copyright (C) 2023-2024 NetIn Developers
Fariba Karimi <[email protected]>
Lisette Espin-Noboa <[email protected]>
Jan Bachmann <[email protected]>

networkinequalities's People

Contributors

Stargazers

Watchers

Forkers

sudhang

networkinequalities's Issues

Sphinx configuration error / conflict with dependencies

The documentation cannot be re-build when adding the Sampling classes from the netin.algorithms.sampling module. see deployment error from workflow Documentation #20: Commit 7f0c636]

Citation backlinks overlap with author list

In the citation list, backlinks overlap with the list of authors.

TC not shown in documentation

@mannbach please take a look.
Also applies for g_tc (if necesary).

Implement graph-tool

We should provide an alternative to NetworkX to improve performance.

Some requirements from the top of my head:

Using graph-tool should be optional. Only when people specifically select it, we should require it to be installed.
It should be hidden behind an abstract network layer (which needs to be implemented for this purpose). This abstract layer describes the network and the functions we need (e.g., for adding nodes, labels and analyses). From the abstract network layer, inheriting classes can implement the access to NetworkX and graph-tool objects respectively.

Biased selection of triadic_closure_candidates

Right now, the selection of targets from the set of triadic_closure_candidates is biased towards those with more connections in common (i.e., weighted by friends-in-commom).

We need an alternative version where nodes are selected uniformly at random by adding a flag in the constructor of netin.TriadicClosure

Types of undirected edges

The method netin.stats.networks.get_edge_type_counts should merge the counts for Mm and mM when the network is undirected.

Plotting types of edges

Add a method to plot edge counts in netin.viz.handlers, and include parameter to plot fraction of edges too.

Overall fraction across all edges in the network
Fraction within group: (mm and mM), and (MM and Mm)

The documentation navigation bar overlaps with the contents

In the documentation, some of the items overlapt with the actual content.

Implement binary tree sampling method for preferential attachment

Use a binary tree data structure to store (unnormalized) probabilities for picking nodes with preferential attachment. This avoids computing the whole array of probabilities with each new node.

Instead, one can traverse a binary tree in which the leaves are target nodes. Their parent nodes are defined by the cumulative sum of their two children. One can then flip O(log N) coins (with N existing nodes) in each round to randomly traverse to the bottom and hence find a target to connect to. At each level, the coin is weighted by the cumulative sum of the two children. The tree can be extended by updating only the path to the newly added leave node.

Order in target selection is not deterministic

In the network generation process, when a new node wants to link to an existing once, we collect potential target candidates together with the probability to connect to them.
See, for instance, pah.get_target_probabilities.
The return type is a list of probabilities and a set of target nodes.

The order over which a set iterated, is not defined and can vary from run to run, even when the set content is identical.
We cannot be sure that the first probability in the list belongs to the nodes which appeared first in an iteration over the corresponding set.
This can lead to assigning the wrong connection probability to nodes in random ways.

We should either

unify both in a dictionary or
additionally return an ordered list of candidate nodes or
remove the target set (typically, only the probabilities are extracted)