← Papers
Classical ROM

Deep LSPG

Model reduction on nonlinear manifolds with convolutional autoencoders

Paper title Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders

Lee and Carlberg (Sandia National Laboratories) · Energy & fusion · General PDEs

Paper arXiv 1812.08373PDF

Replaces the linear POD basis of a projection-based reduced-order model with the decoder of a convolutional autoencoder, so the state lives on a curved low-dimensional manifold. The governing equations are still solved at each time step, by Galerkin or least-squares Petrov-Galerkin projection onto that manifold, with a posteriori error bounds.

Published
JCP

Share on LinkedIn, X or Bluesky, or or .

Know related work that belongs here?

Something wrong on this page?

More papers

All papers
  1. Classical ROMPOD-NN — Non-intrusive reduced-order models: POD modes with neural-network coefficients
  2. Classical ROMDMD — Dynamic mode decomposition of numerical and experimental data
  3. GraphsR-BiLoGo — Transfer learning for rib effects in sheet-forming surrogates