"Learning with Statistical Relations" - Seminario Prof. E. Isufi
Il Dipartimento di Ingegneria è lieto di invitarvi al seminario del Prof. Elvin Isufi, della Technical University of Delft (NL) dal titolo Learning with Statistical Relations: Advanced CoVariance Neural Networks, che si terrà presso l'Aula magna del Dipartimento di Ingegneria, lunedì 19 Ottobre 2026 alle ore 12.00.
Il Prof. Isufi è un brillante ricercatore la cui attività scientifica è posizionata all'intersezione di molte discipline che includono signal processing, deep learning, artificial intelligence, statistical learning, e data science.
Questa attività trova le sue radici proprio presso il Dipartimento di Ingegneria, dove è stato studente oltre che Dottorando di ricerca in collaborazione con l'Università di Delft.
Anche per questo motivo, siamo orgogliosi di riaverlo tra noi nella nuova veste di Professore e affermato ricercatore a livello internazionale.
Per motivi organizzativi è gradita la registrazione al seminario, che è disponibile a questo link, e che potrebbe essere utile per ricevere informazioni dell'ultimo momento su eventuali cambi di programma:
Abstract:
Covariance information is commonly used in machine learning to reveal data interdependencies such as network topology inference (e.g., graphical lasso) and dimensionality reduction (e.g., Principal Component Analysis (PCA)). However, such information is often only the first step in a machine learning pipeline that is performed separately from the task. Because of finite-data estimation errors, we end up working with a sample covariance matrix that leads to uncertainties in its spectrum. For example, PCA is notoriously unstable to covariance estimation errors, i.e., small data perturbations might lead to large changes in principal directions. To address this, coVariance Neural Networks (VNNs) were introduced. These networks perform graph convolutions on the sample covariance matrix, an operation that, similarly to PCA, modulates the data principal components, but with enhanced representation power and greater stability against covariance estimation errors. However, in sparse, high-dimensional settings with limited data, covariance estimation is particularly difficult, which hinders VNNs’ performance despite their stability. Sparse VNNs overcome this by using theoretically grounded covariance sparsification, which improves stability, reduces the impact of spurious correlations on performance and improves computation and memory efficiency. The success of VNNs motivates their extension to different settings. SpatioTemporal VNNs, for instance, process multivariate time series by applying graph convolutions on the online estimated covariance and temporal convolutions over time, achieving stability to estimation errors in both covariance and model parameters due to streaming data. Finally, VNNs’ stability promotes fairness in datasets with poorly represented groups. Building on this, Fair VNNs leverage equitable covariance estimates and fairness penalties in the loss function to ensure a more balanced treatment of these groups.
Bio:
Elvin Isufi (Senior Member, IEEE) is an Associate Professor in the Faculty of Electrical Engineering, Mathematics and Computer Science at Delft University of Technology (TU Delft), The Netherlands, where he chairs the Multimedia Computing Section and co-founded AIdroLab, TU Delft’s AI Lab for graph-based learning in water networks. He received his Ph.D. from TU Delft in 2019 and his M.Sc. (‘14) and B.Sc. (‘12) degrees in Electronic and Telecommunication Engineering from the University of Perugia, Italy. His research focuses on the structure and dynamics of relational data, including graphs and topologies, with applications to critical infrastructure networks. His work lies primarily within the scope of the IEEE Signal Processing Society (SPS). His recognitions include the 2022 IEEE SPS Best Ph.D. Dissertation Award and the IEEE ICASSP 2025 Best Conference Paper Award, as well as paper recognitions at IEEE CAMSAP (‘17), DSLW (‘21, ‘22), ICASSP (‘23), and Asilomar (‘25). He has served in several IEEE roles and currently serves as an Associate Editor for the IEEE Transactions on Signal Processing. He previously served as an Associate Editor for Signal Processing (Elsevier, ’24-'26). He is an ELLIS Scholar (‘26), an NWO Veni fellow (‘24), and a Horizon MSCA COFUND fellow (‘19).