Graphia

Introduction

graphia is a tool created for the "Machine Learning For Network Modeling" module of the "parcours IA".

The goal was to measure some properties of time-evolving graphs in order to create a suitable model for such graphs.

The first properties considered are :

• Average degree : the average degree of the node at each time sample
• Inter-contact histogram : the histogram of the time between two node come in contact againp
• Creation/deletion fraction : the number of created/deleted edges over the number of edges that could have been created/deleted

The first model used is the Edge-Markovian model : given a graph at a given time, the structure of the graph at the next time step depends directly on it. The stucture change is ruled by two independent parameters : the probability that a new link is created and the probability that an existing link is deleted.

Installation

Use precompiled binary

Go to the release page of the repository and download the binary corresponding to your system.

Windows

Just run the binary :

graphia.exe --help

Linux

Set the file permissions as executable : chmod +x graphia-linux and run it :

./graphia-linux --help

MacOs

A binary is provided but not tested

Install via Cargo

This is the most reliable method. Install Rust on your machine, clone the repository and run cargo.

Once Rust is installed, run :

git clone https://github.com/grodino/graphia.git
cd graphia
cargo run -- --help

The -- after cargo run are here to pass the command line arguments to graphia after it has been built by cargo.

Examples

The CLI tool is subdivised in commands, you can list them via graphia - (or cargo run -- - if you use cargo).

To see the global help, run graphia --help. To see the help associated to a command, run graphia <command> --help.

From now one, lets assume that you have a folder data at the same level as graphia and that it contains two dataset files: Rollernet and Infocom06

To display the analyse of a dataset an save it in a directory named generated, run :

graphia --save generated/ analyse data/Rollernet

Do not forget the slash at the end of the folder

To compare a dataset and the results of a simulation based on EdgeMarkovian from this dataset, save graphs in folder generated, and not plot the results in new windows, run :

graphia --no-show --save generated/ compare 1 data/Rollernet