Deep Embedded Multiplicative DMD for Algebra-Preserving Koopman Learning

Published
Source
arXiv
Paper number
320
Field
Machine Learning
arXiv ID
2606.05131

Key points

  • Deep Koopman methods learn flexible coordinates, while structure-preserving methods enforce operator identities over a fixed dictionary.
  • The result is a finite transition map over the learned latent cells.
  • These results suggest a practical rule for Koopman learning: learn the coordinates, constrain the algebra.

Paper links

External research summaries. These are not HDATF publications or measured product results.

Read original (opens in a new tab)