MLG workshop supplementary material

This repository contains the source code of CURGRAPH algorithm and of the experiments of the paper "Iterative Radius-Constrained Clustering" accepted in Machine Learning With Graphs Workshop at ECML-PKDD 2025.

Installation

Requirements

• Python 3 (tested with Python >= 3.9)
• GCC (tested with 12.2.0)
• (Optional but recommended) Docker (tested with 20.10.24)

Quick install

Once you have verified that you have the required pieces of software, you can either use docker to build and run the experiments in a container, or you can install and run the experiments directly on your machine.

With Docker

To build the docker image, you can use the following command:

cd <path-to-repository>
docker build . -t mlg-exp

This will build the docker image with the name mlg-exp, ensuring you have the required pieces of software and libraries to run the experiments, and isolating the environment from your machine.

Manual installation

If you prefer to install and configure dependencies manually, please refer to the INSTALL.md file.

USAGE

You can run the experiments with or without Docker, depending on how you installed the dependencies. Each method will run the experiments with the article settings (all datasets, 10 max clusters). The results will be saved in the results folder.

Note

We strongly recommend using Docker to run the experiments, as it will ensure that the environment is correctly set up and that the results are consistent.

Note

In order to get fast result files to dig in, we recommend to run the experiments in the order provided below.

Running with Docker

To run the experiments with Docker, you can use the following commands.

docker run -v $(pwd)/results:/mds_experiments/results mlg-exp:latest general -m <max-clusters> 

If you need help using the CLI, you can use the following command to get the help message:

docker run mlg-exp:latest --help

Running without Docker

python3 experiments/cli.py general -m <max-clusters>

If you need help using the CLI, you can use the following command to get the help message:

python3 experiments/cli.py --help

Results

Upon completion, the results will be saved in the results/<type-of-campaign>/<Date in ISO format>/ folder in three files:

• raw.csv: contains the raw results of the experiments
• bench.csv: contains the results aggregated by dataset and algorithm
• stats.csv: contains the statistical scores of Curgraph on this campaign

License

This project is licensed under the MIT License - see the LICENSE file for details.