The paper “Reduced order modeling with shallow recurrent decoder networks”, signed by Matteo Tomasetto, Jan P. Williams, Francesco Braghin, Andrea Manzoni, and J. Nathan Kutz, has been published in the prestigious journal Nature Communications.
The work proposes a new deep learning architecture (SHRED-ROM) capable of reconstructing high-dimensional state dynamics, such as those arising in chaotic and nonlinear fluid dynamics, across multiple scenarios from the temporal history of limited sensor measurements, with remarkable gains in accuracy and efficiency compared to several strategies recently appearing in the literature.
Link to article: https://www.nature.com/articles/s41467-025-65126-y
