We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Diffeomorphically Learning Stable Koopman Operators

Abstract: System representations inspired by the infinite-dimensional Koopman operator (generator) are increasingly considered for predictive modeling. Due to the operator's linearity, a range of nonlinear systems admit linear predictor representations - allowing for simplified prediction, analysis and control. However, finding meaningful finite-dimensional representations for prediction is difficult as it involves determining features that are both Koopman-invariant (evolve linearly under the dynamics) as well as relevant (spanning the original state) - a generally unsupervised problem. In this work, we present Koopmanizing Flows - a novel continuous-time framework for supervised learning of linear predictors for a class of nonlinear dynamics. In our model construction a latent diffeomorphically related linear system unfolds into a linear predictor through the composition with a monomial basis. The lifting, its linear dynamics and state reconstruction are learned simultaneously, while an unconstrained parameterization of Hurwitz matrices ensures asymptotic stability regardless of the operator approximation accuracy. The superior efficacy of Koopmanizing Flows is demonstrated in comparison to a state-of-the-art method on the well-known LASA handwriting benchmark.
Comments: Revised version submitted to IEEE Control Systems Letters (L-CSS) with substantially revised exposition, evaluation and proof of Lemma 2 (previously Lemma 8)
Subjects: Machine Learning (cs.LG); Systems and Control (eess.SY)
Cite as: arXiv:2112.04085 [cs.LG]
  (or arXiv:2112.04085v2 [cs.LG] for this version)

Submission history

From: Petar Bevanda [view email]
[v1] Wed, 8 Dec 2021 02:40:40 GMT (2385kb,D)
[v2] Mon, 30 May 2022 08:55:58 GMT (6431kb,D)

Link back to: arXiv, form interface, contact.