{"id":"arxiv/2601.11259","url":"https://arxiv.org/abs/2601.11259","tags":["cs.LG","math.NA"],"score":0,"title":"Latent Dynamics Graph Convolutional Networks for model order reduction of parameterized time-dependent PDEs","domain":"cs.LG","source":"arxiv","authors":["Lorenzo Tomada","Federico Pichi","Gianluigi Rozza"],"reviews":0,"subject":"cs.LG","summary":"Graph Neural Networks (GNNs) are emerging as powerful tools for nonlinear Model Order Reduction (MOR) of time-dependent parameterized Partial Differential Equations (PDEs). However, existing methodologies struggle to combine geometric inductive biases with interpretable latent behavior, overlooking dynamics-driven features or disregarding spatial information. In this work, we address this gap by introducing Latent Dynamics Graph Convolutional Network (LD-GCN), a purely data-driven, encoder-free architecture that learns a global, low-dimensional representation of dynamical systems conditioned…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-01-16","reactions":{}}