Convolutional Symmetric AutoEncoders: enhancing latent stability via differential geometry

DOI10.48550/arXiv.2607.00669MaRDI QIDQ7270498zbMATH OpenarXivFDO

Authors L. Magri, Gianluigi Rozza, G. Li Causi, N. Tonicello, G. Rozza

Publication date 1 July 2026

Published in arXiv

Tldr Convolutional Symmetric AutoEncoders (CAEs) are introduced to improve dimensionality reduction in Reduced-Order Modelling of parametric PDEs by preserving geometric properties of the solution manifolds. CAEs significantly outperform classical architectures on three benchmark problems, exhibiting improved predictive capabilities and robustness.

Full work available at URL https://doi.org/10.48550/arXiv.2607.00669

arXiv classification

math.NAstat.ML

Cites work

• A comprehensive deep learning-based approach to reduced order modeling of nonlinear time-dependent parametrized PDEs
• A graph convolutional autoencoder approach to model order reduction for parametrized PDEs
• Attractors for equations of mathematical physics
• Breaking the Kolmogorov Barrier with Nonlinear Model Reduction
• Data-driven forecasting of high-dimensional chaotic systems with long short-term memory networks
• Deep convolutional recurrent autoencoders for learning low-dimensional feature dynamics of fluid systems
• Deep Symmetric Autoencoders from the Eckart-Young-Schmidt Perspective
• Dynamic mode decomposition of numerical and experimental data
• Equivariant and coordinate independent convolutional networks. A gauge field theory of neural networks
• Fourth-Order Time-Stepping for Stiff PDEs
• In What Sense is the Kolmogorov-Sinai Entropy a Measure for Chaotic Behaviour?—Bridging the Gap Between Dynamical Systems Theory and Communication Theory
• Inertial manifolds
• Inertial manifolds and finite-dimensional reduction for dissipative PDEs
• Inertial manifolds for nonlinear evolutionary equations
• Learning Nonlinear Projections for Reduced-Order Modeling of Dynamical Systems using Constrained Autoencoders
• Long-time prediction of nonlinear parametrized dynamical systems by deep learning-based reduced order models
• Lyapunov characteristic exponents for smooth dynamical systems and for Hamiltonian systems; a method for computing all of them. I: Theory
• LYAPUNOV EXPONENTS OF THE KURAMOTO–SIVASHINSKY PDE
• Multiscale graph neural network autoencoders for interpretable scientific machine learning
• Neural Ordinary Differential Equations
• Non-linear manifold reduced-order models with convolutional autoencoders and reduced over-collocation method
• On latent dynamics learning in nonlinear reduced order modeling
• Predicting turbulent dynamics with the convolutional autoencoder echo state network
• Reduced order methods for modeling and computational reduction
• Smooth manifolds
• Stability analysis of chaotic systems from data
• The Kolmogorov N-width for linear transport: Exact representation and the influence of the data
• The Lyapunov Characteristic Exponents and their computation
• Turbulent Flows
• Using machine learning to replicate chaotic attractors and calculate Lyapunov exponents from data

This page was built for publication: Convolutional Symmetric AutoEncoders: enhancing latent stability via differential geometry

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7270498)