Dipartimento di Ingegneria informatica, automatica e gestionale

Deep Learning

• Graph Neural Networks (GNNs): Exploration of spectral and topological GNNs, as well as Graph Transformer Networks (GTNs) for complex graph-based tasks. 
• Physics-Inspired Machine Learning and Geometric Deep Learning: Integration of physical principles and geometric biases into machine learning frameworks for enhanced interpretability, regularization, and data-efficiency.
• Explainable Machine Learning: Creating interpretable models and tools to enhance transparency and trust in AI systems.

Uncertainity Modeling

• Bayesian Modeling: Development and application of probabilistic models, including hierarchical Bayesian models, for uncertainty quantification and decision-making. 
• Learning with Positive-Unlabeled and Biased Data: Designing robust, generalizable algorithms and statistical techniques for effective learning under partial labeling constraints and and observational biases.
• Quantifying Uncertainty in Machine Learning Predictions: Developing rigorous methodologies to quantify, assess, and address uncertainty in predictions, enhancing model reliability and decision-making under uncertainty.

Complex Systems and Large-scale Complex Networks:

• Complex Systems Analysis: Investigating emergent dynamics and interactions within interconnected systems, with a focus on ecological systems and bipartite interaction networks.
• Link Prediction: : Developing predictive methodologies to tackle challenges of link-predictions in heterogeneous and incomplete interaction graphs.
• Ecopidemiological Modeling: Investigating the spread of different viral families across animal species leveraging phylogenetic and biogeographic patterns, along with emergent dynamics from known host-virus interactions.