git-artes / cpd_rdpg

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

cpd_rdpg: Change-Point Detection on Random-Dot Product Graphs (RDPG)

A dynamic graph showing matches between national football teams.

In this repository we share (and keep) implementations of some algorithms regarding the problem of Change-Point Detection (CPD) on graphs. We are particularly interested in the very versatile RDPG model for random graphs (see for instance, this very complete survey). As of this moment, we have two algorithms:

  • The offline CPD algorithm described in ``Change point localization in dependent dynamic nonparametric random dot product graphs'' by Oscar Hernan Madrid Padilla, Yi Yu, Carey E. Priebe (preprint available at https://arxiv.org/abs/1911.07494). This is implemented in cpd.py.
  • The online CPD algorithm described in ``Online Change Point Detection for Random Dot Product Graphs'' by Bernardo Marenco, Paola Bermolen, Marcelo Fiori, Federico Larroca and Gonzalo Mateos. Available soon. This is implemented in cpd_online.py.

It depends on libraries typically available on most modern Python distributions (in particular NetworkX), with the exception of Graspologic. Visit https://graspologic.readthedocs.io/.

We have included the main modules and some simple examples we've used to generate some of the graphs of the associated papers. Additionally, we have included some examples that use real datasets and both online and offline methods:

About

License:GNU General Public License v3.0


Languages

Language:Python 100.0%