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

Download:

Current browse context:

math.PR

Change to browse by:

References & Citations

Bookmark

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

Mathematics > Probability

Title: Discovering transition phenomena from data of stochastic dynamical systems with Lévy noise

Abstract: In the last decade, many authors have begun to focus on a data-driven approach to analyzing complex dynamics. An advantage of finding dynamic behaviors from data is that this approach enables the investigation of nonlinear phenomena whose mathematical models are unavailable. The purpose of this present work is to propose a framework for quantifying transition phenomena, such as mean exit time and escape probability, for stochastic differential equations from data. As a tool in describing dynamical systems, the Koopman operator transforms a nonlinear system into a linear system, but at the cost of elevating a finite dimensional problem into an infinite dimensional one. In spite of this, using the relation between the stochastic Koopman operator and the infinitesimal generator of a stochastic differential equation, we propose a data-driven method to calculate the mean exit time and escape probability in transition phenomena. Specifically, we use the finite dimensional approximation of the Koopman operator by extending a dynamic mode decomposition algorithm, to obtain the finite dimensional approximation of the infinitesimal generator. We then obtain the drift term and diffusion term for the stochastic differential equation. Finally, we compute the mean exit time and escape probability. This framework is applicable to extract transition information from data of stochastic differential equations with either (Gaussian) Brownian motion or (non-Gaussian) L\'evy motion. We present one - and two-dimensional examples to demonstrate the effectiveness of our framework.
Comments: 21 pages, 12 figures
Subjects: Probability (math.PR); Dynamical Systems (math.DS)
Cite as: arXiv:2002.03280 [math.PR]
  (or arXiv:2002.03280v1 [math.PR] for this version)

Submission history

From: Yubin Lu [view email]
[v1] Sun, 9 Feb 2020 04:41:33 GMT (2347kb,D)

Link back to: arXiv, form interface, contact.