Datum predavanja: utorak, 31.3.2026.
Vrijeme predavanja: 12h u B18

Juan Ricardo Muñoz Galeano, UNIDU
Title: Reduced-Order Modeling for Parameter-Dependent Optimal Control: A map-parameter-to-latent approach

Abstract:
The optimal control of parameter-dependent partial differential equations (PDEs) arises in a wide range of scientific and engineering applications. Classical discretization techniques lead to large-scale dynamical systems whose computational cost becomes prohibitive when multiple evaluations over a parameter space are required. Reduced order models (ROMs) aim to overcome this limitation by constructing low-dimensional approximations of the solution manifold. However, classical projection-based approaches may fail in transport-dominated regimes due to the slow decay of Kolmogorov $n$-widths. In this work, we aim to implement a nonlinear framework based on autoencoders combined with parameter-to-latent mappings to solve efficiently the optimal control problem.