GIMC SIMAI Young 2026

Session

MS02.2 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

4 Jun 2026, 16:30

Pisa

77. Deep Symmetric Autoencoders: Theoretical Framework and Practical Insights

Simone Brivio (MOX, Dipartimento di Matematica, Politecnico di Milano)

04/06/2026, 16:30

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

Nowadays, autoencoders play a major role in reduced order modeling as they allow to represent the solution manifold of complex parameterized PDEs using few latent coordinates, thanks to their nonlinear nature. Indeed, their expressive power enables them to overcome the limitations of linear reduction methods, such as proper orthogonal decomposition (POD), which cannot break the so-called...

81. Multi-scale simulations of excitable cells through Finite-Basis PINNs

Luca Pellegrini (University of Pavia)

04/06/2026, 16:45

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

Physics-Informed Neural Networks (PINNs) have emerged as a powerful tool for solving forward and inverse problems by embedding physical laws directly into the loss function. However, PINNs often struggle with spectral bias, making it difficult to capture the high-frequency oscillations and stiff dynamics present in modeling excitable cells. Finite-Basis PINNs (FBPINNs) address these...

87. Bridging Numerical Methods and Neural Networks for PDEs

Giovanni Pagano (Department of Agricultural Sciences, University of Naples Federico II, Italy)

04/06/2026, 17:00

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

Partial Differential Equations (PDEs) play a central role in modeling complex phenomena arising in diverse applications, including battery life cycles, vegetation dynamics, renewable energy systems. Alongside classical numerical discretization techniques, recent advances increasingly rely on Neural Networks, particularly Physics-Informed Neural Networks (PINNs) [1], which approximate PDEs...

98. Physics Informed Neural networks for downscaling on irregular meshes: successes and limitations

Nemo Malhomme (Sant'Anna Pisa)

04/06/2026, 17:15

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

The DANTE project aims to create computationally efficient models of urban microclimate by applying model order reduction techniques to high-resolution urban-scale simulations. Resulting models must undergo a rigorous validation process before any application is possible, to ensure their accuracy and quantify their uncertainties. This validation process requires urban-scale ground truth data,...

104. Leveraging neural active manifolds for stratified sampling

Andrea Zanoni (Scuola Normale Superiore)

04/06/2026, 17:30

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

Propagating uncertainty from a potentially large number of random inputs through a computational model is becoming increasingly challenging due to the high cost of evaluating complex simulations. Stratified sampling is a well-known variance reduction strategy that, however, has mainly been employed in low-dimensional applications because of the difficulty of extending it to high-dimensional...

111. Generative A.I. meets Reduced-Order Modeling: partial knowledge and hidden parameters in physical systems

Nicola Rares Franco (MOX, Dipartimento di Matematica, Politecnico di Milano)

04/06/2026, 17:45

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

We introduce a new class of generative deep learning based reduced-order models (DL-ROMs) for uncertainty quantification and data-driven modeling of complex physical systems with hidden physics and/or partially observed parameters. Indeed, while DL-ROMs have been extensively shown capable of learning from numerical simulations, existing approaches are predominantly deterministic and assume...

147. A COMPARATIVE BENCHMARK OF FINITE VOLUME, REDUCED ORDER, AND PHYSICS-INFORMED SOLVERS FOR FLOW AROUND CYLINDER

Hassaan Idrees (IMT Lucca)

04/06/2026, 18:00

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

Accurate simulation of unsteady fluid dynamics is critical for applications ranging from aerospace engineering to climate modelling. However, the prohibitive computational cost of high-fidelity solvers often precludes their use in real-time control. Traditionally, Finite Volume Method (FVM) solvers, such as Open FOAM, haveserved as the gold standard for accuracy but they are highly dependent...

163. Machine Learning Surrogates for Robust Inverse Design of Shape-Morphing Elements

Silvia Monchetti (Università degli studi di Firenze - Dipartimento di Ingegneria Civile e Ambientale)

04/06/2026, 18:15

MS02 - Advances in Neural Network Approximation and Surrogate Modeling for Scientific Machine Learning

The inverse design of shape-morphing structures based on responsive polymeric materials, requires the development of theoretical and computational approaches capable of reproducing the involved physical phenomena. When gel-based morphing elements are considered, shape change capabilities can be easily obtained by harnessing their capability to react to external stimuli, such as humidity,...