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: Extended dynamic mode decomposition with dictionary learning using neural ordinary differential equations

Abstract: Nonlinear phenomena can be analyzed via linear techniques using operator-theoretic approaches. Data-driven method called the extended dynamic mode decomposition (EDMD) and its variants, which approximate the Koopman operator associated with the nonlinear phenomena, have been rapidly developing by incorporating machine learning methods. Neural ordinary differential equations (NODEs), which are a neural network equipped with a continuum of layers, and have high parameter and memory efficiencies, have been proposed. In this paper, we propose an algorithm to perform EDMD using NODEs. NODEs are used to find a parameter-efficient dictionary which provides a good finite-dimensional approximation of the Koopman operator. We show the superiority of the parameter efficiency of the proposed method through numerical experiments.
Comments: Corrigendum: The loss function in Eq. (20) is not what we have used in our code. Please replace the sum of squared error in Eq. (20) with the mean squared error
Subjects: Machine Learning (cs.LG); Signal Processing (eess.SP); Dynamical Systems (math.DS); Numerical Analysis (math.NA); Chaotic Dynamics (nlin.CD); Data Analysis, Statistics and Probability (physics.data-an)
Journal reference: Nonlinear Theory and Its Applications, IEICE, vol. 12, no. 4, pp. 626-638, 2021
DOI: 10.1587/nolta.12.626
Cite as: arXiv:2110.01450 [cs.LG]
  (or arXiv:2110.01450v2 [cs.LG] for this version)

Submission history

From: Sho Shirasaka [view email]
[v1] Fri, 1 Oct 2021 06:56:14 GMT (2288kb)
[v2] Tue, 17 May 2022 04:53:20 GMT (2288kb)

Link back to: arXiv, form interface, contact.