INDRO SPINELLI

Dottore di ricerca

ciclo: XXXV


Titolo della tesi: Towards Trustworthy Graph Neural Networks

This thesis aims to develop and deploy deep learning algorithms for graph-structured data people trust for real-world applications. To be trusted from a human perspective requires robustness from malign input manipulation, privacy protection of the ingested data, fair treatment for every individual or group of individuals in the data and transparency in the decision process. However, the ``vanilla'' formulation of graph neural networks(GNNs) do not respect all these characteristics. We address these issues one by one through a series of different works. First, we propose solutions to create robust architectures for the dataset at hand and fully distributed GNNs to preserve data privacy. We put an accent on the fairness of the predictions by removing the bias directly from the data. In particular, the tendency of similar nodes to cluster on several real-world graphs (i.e., homophily) can dramatically worsen the fairness of these procedures. First, we propose a biased pruning of the graph connections to reduce the homophily of the sensitive traits. Secondly, instead of dropping edges at random, we learn a new and fairer version of the graph's topology. Finally, we tried associating additional information with the GNNs' predictions to allow human experts to interpret and extract knowledge from the model. We develop a meta-learning framework for improving the level of explainability of a GNN at training time by steering the optimization process towards an ``interpretable'' local minima. Then, we propose an architecture trained over a bag of explanation subgraphs used to improve prediction performances and constitute an easy-to-interpret explanation. To conclude, we present a real user case in industrial settings where GNN can be combined with Convolutional Neural Networks to re-identify objects from aerial photos, potentially improving the quality of service of millions of people.

Produzione scientifica

11573/1702063 - 2024 - ArcheoWeedNet: Weed Classification in the Parco archeologico del Colosseo
Saurio, G.; Muscas, M.; Spinelli, I.; Rughetti, V.; Della Giovampaola, I.; Scardapane, S. - 04b Atto di convegno in volume
congresso: Proceedings of the 22nd International Conference on Image Analysis and Processing, ICIAP 2023 (Udine; Italia)
libro: Image Analysis and Processing - ICIAP 2023 Workshops - (9783031510229; 9783031510236)

11573/1702039 - 2023 - Explainability in subgraphs-enhanced Graph Neural Networks
Guerra, Michele; Spinelli, Indro; Scardapane, Simone; Bianchi, Filippo Maria - 04b Atto di convegno in volume
congresso: Northern Lights Deep Learning Workshop 2023 (Tromso; Norway)
libro: Vol. 4 (2023): Proceedings of the Northern Lights Deep Learning Workshop 2023 - ()

11573/1693646 - 2023 - Interpreting black-box models. A review on explainable artificial intelligence
Hassija, V.; Chamola, V.; Mahapatra, A.; Singal, A.; Goel, D.; Huang, K.; Scardapane, S.; Spinelli, I.; Mahmud, M.; Hussain, A. - 01a Articolo in rivista
rivista: COGNITIVE COMPUTATION (New York : Springer) pp. 1-30 - issn: 1866-9964 - wos: WOS:001060142200001 (7) - scopus: 2-s2.0-85168561911 (15)

11573/1695410 - 2023 - ICML 2023 topological deep learning challenge. Design and results
Papillon, Mathilde; Hajij, Mustafa; Frantzen, Florian; Hoppe, Josef; Jenne, Helen; Mathe, Johan; Myers, Audun; Papamarkou, Theodore; Schaub, Michael T.; Zamzmi, Ghada; Birdal, Tolga; Dey, Tamal; Doster, Timothy; Emerson, Tegan H.; Gopalakrishnan, Gurusankar; Govil, D.; Grande, Vincent P.; Guzm'an-S'aenz, Aldo; Kvinge, Henry; Livesay, Neal; Meisner, Jan; Mukherjee, Soham; Samaga, Shreyas N.; Natesan Ramamurthy, Karthikeyan; Reddy Karri, Maneel; Rosen, Paul; Sanborn, Sophia; Scholkemper, Michael; Walters, Robin; Agerberg, Jens; Bokman, Georg; Barikbin, Sadrodin; Battiloro, Claudio; Bazhenov, Gleb; Bern('A)Rdez, Guillermo; Brent, Aiden; Escalera, Sergio; Fiorellino, Simone; Gavrilev, Dmitrii; Hassanin, Mohammed; Hausner, Paul; Hoff Gardaa, Odin; Khamis, Abdelwahed; Lecha, M; Magai, German; Malygina, Tatiana; Melnyk, Pavlo; Ballester, Rub('E)N; Varma Nadimpalli, Kalyan; Nikitin, Alexander; Rabinowitz, Abraham; Salatiello, Alessandro; Scardapane, Simone; Scofano, Luca; Singh, Suraj; Sjolund, Jens; Snopov, Paul; Spinelli, Indro; Telyatnikov, Lev; Testa, Lucia; Yang, Maosheng; Yue, Yixiao; Zaghen, Olga; Zia, Ali; Miolane, Nina - 04b Atto di convegno in volume
congresso: 2nd Annual Workshop on Topology, Algebra, and Geometry in Machine Learning, TAG-ML 2023, held at the International Conference on Machine Learning, ICML 2023 (Honolulu; Hawaii)
libro: Proceedings of Machine Learning Research - ()

11573/1693643 - 2023 - Machine Un-learning. An overview of techniques, applications, and future directions
Sai, S.; Mittal, U.; Chamola, V.; Huang, K.; Spinelli, I.; Scardapane, S.; Tan, Z.; Hussain, A. - 01a Articolo in rivista
rivista: COGNITIVE COMPUTATION (New York : Springer) pp. 1-25 - issn: 1866-9964 - wos: WOS:001094347900001 (0) - scopus: 2-s2.0-85175606672 (0)

11573/1695338 - 2023 - Drop edges and adapt. A fairness enforcing fine-tuning for graph neural networks
Spinelli, I.; Bianchini, R.; Scardapane, S. - 01a Articolo in rivista
rivista: NEURAL NETWORKS (Elsevier Science Limited:Oxford Fulfillment Center, PO Box 800, Kidlington Oxford OX5 1DX United Kingdom:011 44 1865 843000, 011 44 1865 843699, EMAIL: asianfo@elsevier.com, tcb@elsevier.co.UK, INTERNET: http://www.elsevier.com, http://www.elsevier.com/locate/shpsa/, Fax: 011 44 1865 843010) pp. 159-167 - issn: 0893-6080 - wos: WOS:001076417000001 (0) - scopus: 2-s2.0-85169037129 (0)

11573/1702188 - 2023 - Combining Stochastic Explainers and Subgraph Neural Networks can Increase Expressivity and Interpretability
Spinelli, Indro; Guerra, Michele; Bianchi, Filippo Maria; Scardapane, Simone - 04b Atto di convegno in volume
congresso: 31st European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (Bruges, Belgium)
libro: ESANN 2023- Proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning - (978-2-87587-088-9)

11573/1671578 - 2022 - Re-identification of objects from aerial photos with hybrid siamese neural networks
Devoto, A.; Spinelli, I.; Murabito, F.; Chiovoloni, F.; Musmeci, R.; Scardapane, S. - 01a Articolo in rivista
rivista: IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS (Institute of Electrical and Electronics Engineers.IEEE) pp. 1-9 - issn: 1551-3203 - wos: WOS:000965676800001 (1) - scopus: 2-s2.0-85133733758 (1)

11573/1671581 - 2022 - A meta-learning approach for training explainable graph neural networks
Spinelli, I.; Scardapane, S.; Uncini, A. - 01a Articolo in rivista
rivista: IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS (Piscataway, NJ : Institute of Electrical and Electronics Engineeers) pp. 1-9 - issn: 2162-237X - wos: WOS:000795116700001 (4) - scopus: 2-s2.0-85132538173 (3)

11573/1486181 - 2021 - Distributed training of graph convolutional networks
Scardapane, S; Spinelli, I; Di Lorenzo, P - 01a Articolo in rivista
rivista: IEEE TRANSACTIONS ON SIGNAL AND INFORMATION PROCESSING OVER NETWORKS (New York, NY : IEEE Signal Processing Society [u.a.]) pp. 87-100 - issn: 2373-776X - wos: WOS:000608678000001 (10) - scopus: 2-s2.0-85098777287 (9)

11573/1476980 - 2021 - Efficient data augmentation using graph imputation neural networks
Spinelli, I.; Scardapane, S.; Scarpiniti, M.; Uncini, A. - 02a Capitolo o Articolo
libro: Progresses in Artificial Intelligence and Neural Systems - (978-981-15-5092-8; 978-981-15-5093-5)

11573/1612486 - 2021 - Fairdrop. Biased edge dropout for enhancing fairness in graph representation learning
Spinelli, Indro; Scardapane, Simone; Hussain, Amir; Uncini, Aurelio - 01a Articolo in rivista
rivista: IEEE TRANSACTIONS ON ARTIFICIAL INTELLIGENCE (Piscataway NJ: IEEE) pp. 1-11 - issn: 2691-4581 - wos: (0) - scopus: 2-s2.0-85132958155 (15)

11573/1486183 - 2020 - Adaptive propagation graph convolutional network
Spinelli, I; Scardapane, S; Uncini, A - 01a Articolo in rivista
rivista: IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS () pp. 1-6 - issn: 2162-2388 - wos: WOS:000704111000042 (40) - scopus: 2-s2.0-85117164528 (34)

11573/1486185 - 2020 - Missing data imputation with adversarially-trained graph convolutional networks
Spinelli, I; Scardapane, S; Uncini, A - 01a Articolo in rivista
rivista: NEURAL NETWORKS (Elsevier Science Limited:Oxford Fulfillment Center, PO Box 800, Kidlington Oxford OX5 1DX United Kingdom:011 44 1865 843000, 011 44 1865 843699, EMAIL: asianfo@elsevier.com, tcb@elsevier.co.UK, INTERNET: http://www.elsevier.com, http://www.elsevier.com/locate/shpsa/, Fax: 011 44 1865 843010) pp. 249-260 - issn: 0893-6080 - wos: WOS:000555878900001 (60) - scopus: 2-s2.0-85086438644 (67)

© Università degli Studi di Roma "La Sapienza" - Piazzale Aldo Moro 5, 00185 Roma