UNIPD – Italy

The University of Padova (UNIPD), founded in 1222, has about 60000 students, with about 2500 academic staff members, and a total number of about 5000 employees. UNIPD has been recently ranked in a government investigation on the quality of research the top Italian (large) university and is currently involved in a number of EU and nationally funded projects. DM is composed of about 150 academic staff members, carrying out research in all areas of mathematics. The Numerical Modeling group of DM emanates from the union of two analogous at different departments groups after internal reorganization of UNIPD in 2012. As a result of this union, the more theoretical group has been complemented by more applied researchers resulting in a team with a long-standing background in both development and analysis of numerical models and in machine learning algorithms, with strong connections with applications.

The project will essentially be implemented by:

  • Mario Putti – Associate Professor of Numerical Analysis. Expertise in numerical solution of PDEs with applications to simulation of the environmental fluid and structural dynamics of natural resources.
  • Fabio Aiolli – Assistant Professor in Information Science. Expertise in kernel methods for Machine Learning. Kernels for structured data, kernel and representation learning, hierarchical representations and deep learning. Applications to recommender systems, neuroscience and biology.
  • Stefano De Marchi – Associate Professor of Numerical Analysis. Expertise in stability issues of multivariate approximation by polynomials, radial basis functions and rational approximation. Applications to medical image reconstruction by CT/MRI and MPI
  • Alvise Sommariva – Associate Professor of Numerical Analysis. Expertise in in numerical multivariate interpolation and cubature, and their applications
  • Alessandro Sperduti – Full Professor in Information Science. Expertise in in Machine Learning, with focus on (deep) neural networks and kernel methods for structured domains; data and process mining. Applications to business process management, cognitive systems, and biology.
  • Marco Vianello – Associate Professor of Numerical Analysis. Expertise in in multivariate polynomial regression, interpolation and numerical cubature with applications

Key expertise

Development and analysis of numerical methods for application to the simulation of problems in groundwater hydrology, surface-subsurface water interaction, flow and transport modelling, geomechanics and land subsidence due to fluid withdrawal, data assimilation for control and identification of hydrological systems, etc. Another strong area of expertise is in the field of multivariate approximation theory, numerical linear algebra, and machine learning. This expertise culminated in the development of several libraries for High Performance Computing. Finally, strong expertise in kernel methods for machine learning is present in the research group. This expertise will be essential to successfully pursue the ideas set forth in this proposal. The group has a strong past experience in the participation and coordination of large scale projects (funded by EU, national agencies, private companies).

Together with Claudio Paniconi of INRS-Quebec, Canada, Mario Putti, the group coordinator, has been the original developer of the CATHY (CATchment HYdrology) model that will be used within the current project. This model couples surface-subsurface flow and simulates the interactions between soil-plant-atmosphere, a central topic of the proposed project. CATHY has been used in a number of hydrologic applications (most recently the EU projects CLIMB and GLOBAQUA and the Integrated Hydrologic Model Intercomparison Project) and is currently in use at several institutions worldwide, including the LEO (Land Evolution Observatories) at the Biosphere2 project run by the University of Arizona (USA).