Il seminario si svolge martedì 16 aprile, presso la sede CNR-IAC di Bari, in modalità mista, presenza e streaming (vedi link in calce).
Il titolo del seminario è: Physics informed Neural Networks for inverse problems in peridynamics and porous media.
Di seguito l'abstract.
Deep learning is a powerful tool for solving data driven differential problems and has come out to have successful applications in solving direct and inverse problems described by PDEs. When the physics, modeled by a PDE, is fed into a neural network architecture, it gives rise to the so-called Physics Informed Neural Networks (PINNs). In this talk, I will apply PINNs to inverse problems in peridynamic theory, mobile-immobile and generalized multi-rate transfer models and optimal control problems in the framework of Richards’ equation. In particular, I will show results about PINNs, applied to inverse problems in standard diffusion equations, based on Radial Basis Function activated layers; adaptive PINNs for multi-scale models in fluid dynamics; and PINNs used to predict control on moisture content in 2D problems described by Richards’ equation in unsaturated soils.