POD-NN
Non-intrusive reduced-order models: POD modes with neural-network coefficients
Paper title Non-intrusive reduced order modeling of nonlinear problems using neural networks
Paper doi.org
Compresses a set of high-fidelity solutions into a few POD modes, then trains a small two-hidden-layer MLP to map the input parameters (physical or geometric) straight to the mode coefficients. A new case is the network's output times the modes, with no call to the solver; network size and number of training samples are chosen automatically offline using Latin hypercube sampling.
- Published
- JCP
Share on LinkedIn, X or Bluesky, or or .
Know related work that belongs here? Suggest a link
Something wrong on this page? Tell the editor