Numerical and matrix methods for the discrete solution of differential problems and data science

Analysis and development of second order optimization methods for the numerical treatment of problems stemming from
the discretization of PDE-constrained optimal control problems with box constraints.
Special attention to acceleration techniques (preconditioners) for the associated algebraic linear systems,
with focus on optimality properties with respect to the problem parameters (anisotropic coefficients, discretization
mesh, solution regularity, etc.).

Analysis and development of efficient numerical linear algebra methods for the solution of large algebraic linear systems originating
from the discretization of possibly parametrized (systems of) partial differential equations (PDEs), including structured problems,
which may give rise to matrix and tensor equations. Numerical solution of nonlinear matrix equations stemming from evolutive and
parametric PDEs, and from optimal control problems.

Development of matrix and tensor algorithms for the numerical treatment of data science problems, such as object classification,
pattern recognition and information retrieval.

Faculty:

Davide Palitta

Senior assistant professor (fixed-term)

Valeria Simoncini

Full Professor

Research fellows - Ph.D. Students