Research resources
Code and data from our papers, and the portals of projects we take part in. Each entry says what the resource is for and where to find it.
Food Connections: Data and Simulation Tools Project resources on international trade, food and nutrition security, including entry points to data, simulation code and an early-warning application.
Open the resource pageAbout the resource
Food Connections studies how international trade and the rules governing it relate to food and nutrition security and environmental sustainability. The project portal brings together materials and results, with entry points to data, simulation code and an early-warning application.
Use the resource
Consult the linked resources for documentation, data provenance and citation requirements.
Network Dismantling Review: Code and Data Code, benchmark networks and reproducibility material associated with the 2024 review on robustness and resilience of complex networks.
Open the resource pageAbout the resource
This repository accompanies the review:
O. Artime, M. Grassia, M. De Domenico, J. P. Gleeson, H. A. Makse, G. Mangioni, M. Perc, and F. Radicchi. Robustness and resilience of complex networks. Nature Reviews Physics 6, 114–131 (2024).
It brings together implementations and wrappers for network dismantling methods, benchmark networks used in the review and scripts for reproducing comparative results. Network dismantling studies how removing selected nodes or edges changes the connectivity of a network.
Use the resource
Installation and execution instructions are maintained in the repository so that they remain aligned with the code.
Citation and licence
Please cite the review and the original publications for any individual algorithm used. Code produced for the review is released under GPLv3; individual algorithm wrappers can retain the licences of their original implementations. Consult the repository before redistribution or reuse.
CoreGDM: Network Decycling and Dismantling A trainable dismantling algorithm that attacks the 2-core of a network, available in the Network Dismantling review package.
Open the resource pageAbout the resource
CoreGDM was introduced in:
M. Grassia and G. Mangioni. CoreGDM: Geometric Deep Learning Network Decycling and Dismantling. Complex Networks XIV (2023).
It merges the GDM framework with the CoreHD algorithm: it attacks the 2-core of the network, using a learned score in place of the node degree.
Use the resource
The CoreGDM repository points to the review package for now; the stand-alone code is not published yet.
Citation and licence
Please cite the paper. The code is released under GPLv3.
wsGAT: Weighted and Signed Graph Attention A graph attention layer for networks with weighted and signed links, from the 2022 paper on link prediction.
Open the resource pageAbout the resource
This repository accompanies the paper:
M. Grassia and G. Mangioni. wsGAT: Weighted and Signed Graph Attention Networks for Link Prediction. Complex Networks & Their Applications X (2022).
wsGAT extends the Graph Attention Network (GAT) layer to graphs whose links carry a weight and a sign, such as trust and correlation networks. The paper uses it to predict the weight, the sign and the existence of links.
Use the resource
The repository contains the layer,
wsGATConv, written for PyTorch Geometric.Citation and licence
Please cite the paper. The code is released under GPLv3.
GDM: Graph Dismantling with Machine Learning Code, training and test networks, and pre-trained models from the 2021 paper on machine learning dismantling and early-warning signals.
Open the resource pageAbout the resource
This repository accompanies the paper:
M. Grassia, M. De Domenico, and G. Mangioni. Machine learning dismantling and early-warning signals of disintegration in complex systems. Nature Communications 12, 5190 (2021).
A graph neural network trained to dismantle small synthetic networks is used to attack large real-world ones, and its predictions give an early-warning signal of the network’s collapse. The repository holds the scripts, the networks used for training and testing, and the pre-trained models.
Use the resource
Setup and reproduction instructions are in the repository. Only part of the real-world test networks is included, because the original data come under different licences.
Citation and licence
Please cite the paper, and the archived release when you use the code. The code is released under GPLv3.