Inf. Max. with ML methods for Multilayer Networks (TopSpreadersNetwork)

A repository to train and evaluate Influence Maximisation ML models for multilayer networks. This code was used in the preparation of the paper Identifying Super Spreaders in Multilayer Networks.

• Authors: Michał Czuba, Mateusz Stolarski, Adam Piróg, Piotr Bielak, Piotr Bródka
• Affiliation: WUST, Wrocław, Lower Silesia, Poland

Functionality

This repository is part of a broader research codebase composed of multiple interrelated components, each addressing a specific aspect of the Influence Maximisation pipeline:

I. infmax-trainer-icm-mln - training ts-net.
II. infmax-simulator-icm-mln - computing spreading potential and evaluating influence maximisation methods.
III. top-spreaders-dataset - storage and access layer for the TopSpreadersDataset.

Particularly, it contains an implementation, weights, and the training pipeline of the TopSpreadersNetwork (a.k.a. ts-net), which predicts spreading potentials of actors from the processed multilayer network. Its architectural design is presented on the following figure:

Runtime configuration

I. First, initialise the environment:

conda env create -f env/conda.yaml
conda activate infmax-trainer-icm-mln
pip install pyg-lib -f https://data.pyg.org/whl/torch-2.3.1+cu121.html

II. Then, pull the Git submodule with data loaders and install its code:

git submodule init && git submodule update
pip install -e data

III. The TopSpreadersDataset is managed using DVC. To fetch it, follow the instructions in README.md.

Structure of the repository

├── data                    -> evaluated networks
├── env                     -> definition of the runtime environment
├── model                   -> exported model weights & configuration
├── scripts                 -> pipeline entries
│   ├── configs
│   └── analysis
├── src                     -> main source code
│   |── data_models         -> customised HeteroData class
│   |── datamodule          -> data loaders
│   |── dataset             -> implemented datasets serving HeteroData
│   |── infmax_models       -> trainable ML models for super-spreaders ident.
│   |── training            -> training pipeline
│   │   ├── loss
│   │   ├── callbacks.py
│   │   ├── loggers.py
│   │   └── trainer.py
│   |── utils               -> code helpers
│   └── wrapper             -> wrappers for trainable models
├── README.md
├── run_evalation.py        -> entrypoint of evaluation pipeline
└── run_experiments.py      -> entrypoint of training pipeline

Using the package

To run experiments, execute run_experiments.py or run_evaluation.py and provide the appropriate configuration in scripts/configs directory. See examples in scripts/configs for reference.

Running the training pipeline

To run experiments execute: run_experiments.py and provide proper CLI arguments defined in scripts/configs/hydra.yaml, i.e. a name of the configuration file.

To select device on remote server please set up environment variable: export CUDA_VISIBLE_DEVICES=2 and then in the config file select list of devices as [0].

To train model without neptune.ai set up tensor_board logger in configuration file.

Running the evaluation pipeline

To run evaluation execute: run_evaluation.py and provide proper CLI arguments defined in scripts/configs/evaluation.yaml, i.e. a name of the experiment or test networks.

To run it without access to neptune.ai, set the value of base/neptune to False. This will enforce the local configuration from the model directory.

Acknowledgment

This work was supported by the National Science Centre, Poland [grant no. 2022/45/B/ST6/04145] (www.multispread.pwr.edu.pl); the Polish Ministry of Science and Higher Education programme “International Projects Co-Funded”; and the EU under the Horizon Europe [grant no. 101086321]. Views and opinions expressed are those of the authors and do not necessarily reflect those of the funding agencies.