Learning on and with Graphs
Graph structure can guide learning rather than serve merely as another input representation. My work investigates how graph-aware models can support difficult combinatorial problems and how their decisions can be inspected and evaluated.
This includes learned pruning and sub-network extraction for maximum clique enumeration, efficient PageRank improvement through link placement, and graph-neural architectures for multilayer, weighted, and signed networks. Related work addresses explainability and interpretability on graphs, together with evaluation methods for graph generators that go beyond aggregate summary statistics.
Implementations of mGNN and wsGAT are available in the mGNN and wsGAT repositories.
Selected publications
(2026).
Machine-enhanced reconstruction of functional connectomes unravels discriminative brain sub-systems in health and disease.
SciRep.
(2024).
Edge Dismantling with Geometric Reinforcement Learning.
Complex Networks XV.
(2023).
CoreGDM: Geometric Deep Learning Network Decycling and Dismantling.
Complex Networks XIV.
(2022).
Geometric Deep Learning Graph Pruning to Speed-Up the Run-Time of Maximum Clique Enumerarion Algorithms.
Complex Networks XI.
(2022).
Efficient Node PageRank Improvement via Link Building using Geometric Deep Learning.
ACM TKDD.
(2021).
Machine learning dismantling and early-warning signals of disintegration in complex systems.
Nat Commun 12, 5190 (2021).
(2021).
mGNN: Generalizing the Graph Neural Networks to the Multilayer Case.
Submitted to the IEEE Computer Society.
(2021).
wsGAT: Weighted and Signed Graph Attention Networks for Link Prediction.
International Conference on Complex Networks and Their Applications.