Data science and artificial intelligence

The group focuses on quantitative methodologies for Data Science and Artificial Intelligence, with applications in economics, finance, social sciences, medicine, and imaging science. The tools and concepts used stem from mathematical statistics, physics and statistical mechanics, time series theory, operations research, the theory of random and complex networks, and optimization. The group’s interest is aimed both at the theoretical development of new mathematical, statistical, and computational methods, and at the empirical study of ill-posed inverse problems and socio-economic systems, such as interaction networks between people, companies, or, more generally, agents.

In the field of machine learning, the group is interested in the theoretical and numerical understanding of:

Deep learning interpreted as an inverse problem of statistical mechanics with assigned boundary conditions;

The relationships between deep learning and statistical/econometric methods, both from an inferential standpoint and within the context of hypothesis testing;
The development of Scientific Machine Learning approaches, with a focus on hybrid model-based and data-driven models for ill-posed inverse problems, where mathematical models, physical knowledge, regularization techniques, and machine learning methods are integrated to achieve more robust and reliable reconstructions.

Other important research themes for the group include high-dimensional model inference—including dynamic networks and time-varying parameter time series models—the development of clustering methods, entropic and statistical approaches for predicting and validating interactions, and hypothesis testing on networks.

On the theoretical side, soft computing techniques for decision support under vagueness and uncertainty are also studied, formalizing them through fuzzy logic, alongside computational optimization and regularization techniques.

On the applied side, hybrid model-based and data-driven approaches are investigated in applications such as Electrical Impedance Tomography (EIT), Computerized Tomography (CT), and image and surface reconstruction/restoration problems.