Research projects
My research lies at the intersection of graph theory, graph machine learning and network neuroscience. I develop methods and open-source tools for (temporal) networks and apply them to brain data and to other real-world systems.
Network neuroscience
Temporal networks for brain dynamics
I model fMRI recordings as temporal networks, where nodes are brain regions and edges capture correlations that change over time. I study which temporal properties, such as temporal small-worldness, distinguish real brain dynamics from randomness, and proposed temporal hyperbolic random graphs as null models. The resulting temporal brain networks are openly released as datasets.
Explainable machine learning on brain subnetworks
What happens in the brain when we listen to or watch a story? I train a neural network to classify narratives from dynamic functional connectivity, and use Shapley values to quantify how much each of the Yeo 7 subnetworks contributes to the model's decisions.
Graph machine learning
Graph neural networks for centrality ranking
Exact betweenness centrality is prohibitive on large networks. BRAVA-GNN is a lightweight GNN that ranks nodes by betweenness using multi-hop degree masses as size-invariant features. It works on both directed and undirected graphs and uses 56× fewer parameters than the lightest competing GNN.
Open-source software
Deep learning on graphs in Julia
I co-develop GraphNeuralNetworks.jl, a Julia framework for graph neural networks with GPU support and both Flux and Lux backends. During Google Summer of Code 2023 and Julia Summer of Code 2024 I added support for temporal graphs and temporal GNN layers.
Temporal graph analysis in Julia
TemporalGraphs.jl is a pure-Julia library for temporal graphs: temporal distances and optimal paths, temporal centralities (closeness, betweenness, Katz, PageRank), connectivity, flows, spanners, motifs and randomized models. It downloads public temporal networks (SNAP, SocioPatterns) and converts temporal graphs to GNNGraphs for GraphNeuralNetworks.jl.
Graph optimization in Julia
I am the current maintainer of GraphsOptim.jl, a JuliaGraphs package for graph problems solved with mathematical programming: min-cost flow, assignment, graph matching and graph edit distance, maximum clique, independent set, vertex cover, fractional coloring and shortest paths.
AI for mathematics
AI for open problems in graph theory
In the Graph Theory AI project we explore how large language models can track and support progress on open problems in graph theory: a status-annotated catalogue of graph conjectures, LLM proof attempts checked by an adversarial reviewer model, machine-checkable formalizations in Rocq, and Mathpocalypse, where an open-weight LLM re-checks proofs in published papers.
Publications
Journal papers
"Characterizing Dynamic Functional Connectivity Subnetwork Contributions in Narrative Classification with Shapley Values"
Aurora Rossi, Yanis Aeschlimann, Emanuele Natale, Samuel Deslauriers-Gauthier, Peter Ford Dominey
Network Neuroscience, 2025
"GraphNeuralNetworks.jl: Deep Learning on Graphs with Julia"
Carlo Lucibello, Aurora Rossi
Journal of Machine Learning Research, 2025
"A sensitivity analysis of the Earth for all model: Getting the giant leap scenario with fewer policies"
Pierluigi Crescenzi, Giorgio Gambosi, Lucia Nasti, Emanuele Natale, Aurora Rossi
Journal of Industrial Ecology, 2024
"WorldDynamics.jl: A Julia Package for Developing and Simulating Integrated Assessment Models"
Pierluigi Crescenzi, Emanuele Natale, Aurora Rossi, Paulo Bruno Serafim
Journal of Open Source Software, 2024
"On null models for temporal small-worldness in brain dynamics"
Aurora Rossi, Samuel Deslauriers-Gauthier, Emanuele Natale
Network Neuroscience, 2024
Conference papers
"Degree-Mass Message Passing for Betweenness Ranking in Directed and Undirected Networks"
Justin Dachille, Aurora Rossi, Sunil Kumar Maurya, Frederik Mallmann-Trenn, Xin Liu, Frédéric Giroire, Tsuyoshi Murata, Emanuele Natale
The 35th ACM International Conference on Information and Knowledge Management (CIKM 2026), November 7th - 11th 2026, Rome, Italy
"La vie risquée mais enrichissante d'un doctorant interdisciplinaire: accrochez-vous, une seule publication suffit!"
Sayf Halmi, Aurora Rossi, Frédéric Giroire, Nicolas Nisse, Michele Pezzoni
ALGOTEL 2026, June 2026, Mandelieu-la-Napoule, France
"Une implémentation GPU de la méthode de recherche approximative FlyHash"
Arthur da Cunha, Emanuele Natale, Damien Rivet, Aurora Rossi
Conference on Artificial Intelligence for Defense, DGA Maîtrise de l'Information, November 22nd and 23rd 2023, Rennes, France
"Un framework open-source écrit en Julia pour la modélisation d’évaluation globale intégrée"
Pierluigi Crescenzi, Hicham Lesfari, Emanuele Natale, Aurora Rossi, Paulo Serafim
ROADEF 2023, February 20th - 23rd 2023, Rennes, France
Extended abstracts
"Understanding the Importance of Brain Subnetworks with Shapley Values During Narrative Processing"
Aurora Rossi, Yanis Aeschlimann, Emanuele Natale, Samuel Deslauriers-Gauthier, Peter Ford Dominey
Complex Networks Conference, December 10th and 12th 2024, Istanbul, Turkey
"Temporal Hyperbolic Graphs as Null Models for Brain Dynamics"
Aurora Rossi, Samuel Deslauriers-Gauthier, Emanuele Natale
Complex Networks Conference, November 28th and 30th 2023, Menton, France
PhD Thesis
"Computational Methods and Analysis of Temporal Networks: Applications in Neuroscience"
Aurora Rossi
DS4H Université Côte d’Azur, 2025. Special Award for Interdisciplinarity of the École Doctorale STIC.
Posters
"Simulating Global Dynamics with Integrated Assessment Models"
Pierluigi Crescenzi, Aurora Rossi
MOMI2024: Le Monde des Mathematiques Industrielle, April 8th - 9th, Sophia Antipolis, France
"Temporal Graph Neural Networks with GraphNeuralNetworks.jl"
Aurora Rossi
Julia and Optimization Days 2023, October 3rd - 6th, CNAM Paris, France
"Temporal Brain Networks Dataset"
Aurora Rossi, Samuel Deslauriers-Gauthier, Emanuele Natale
NeuroMod Meeting, June 28th - 29th 2023, Antibes, France
"Hyperbolic Model Captures Temporal Small Worldness of Brain Dynamics"
Aurora Rossi, Pierluigi Crescenzi, Samuel Deslauriers-Gauthier, Emanuele Natale
NeuroMod Meeting, June 30th -July 1st 2022, Antibes, France
Datasets
Aurora Rossi, Samuel Deslauriers-Gauthier, Emanuele Natale
(see also here)
Aurora Rossi