Preprint
Jul 2026
Convolutional Symmetric AutoEncoders: enhancing latent stability via differential geometry
This work introduces a novel class of symmetric Convolutional AutoEncoders (CAEs) designed to embody the primary properties of manifold parametrization mappings and demonstrates significantly improved predictive capabilities when integrated into a ROM framework.
Gaspare LI Causi, Niccolò Tonicello, Luca Magri et al.
· 0 citations