Data-Driven Reduction for a Class of Multiscale Fast-Slow Stochastic Dynamical Systems

Abstract

Multiple time scale stochastic dynamical systems are ubiquitous in science and engineering, and the reduction of such systems and their models to only their slow components is often essential for scientific computation and further analysis. Rather than being available in the form of an explicit analytical model, often such systems can only be observed as a data set which exhibits dynamics on several time scales. We will focus on applying and adapting data mining and manifold learning techniques to detect the slow components in such multiscale data. Traditional data mining methods are based on metrics (and thus, geometries) which are not informed of the multiscale nature of the underlying system dynamics; such methods cannot successfully recover the slow variables. Here, we present an approach which utilizes both the local geometry and the local dynamics within the data set through a metric which is both insensitive to the fast variables and more general than simple statistical averaging. Our analysis of the approach provides conditions for successfully recovering the underlying slow variables, as well as an empirical protocol guiding the selection of the method parameters.

Diffusion Maps Kalman Filter for a Class of Systems With Gradient Flows

“…Several studies, e.g. [21], [33] and [34], have shown that based on properties of the diffusion maps coordinates, this assumption commonly holds. Third, we specified the conditions under which the Fig.…”

Intrinsic Isometric Manifold Learning with Application to Localization

“…This approximation has been used is several papers [18,17,16,19,27,37,40,38,39,41] and error analysis for it was presented in [17,37]. The approximation can be viewed as a result of the relation between the intrinsic and observed manifolds.…”

“…The locality of the approximation is demonstrated empirically in section 6. In addition, rigorous error analysis was provided in [17], where it was shown that the approximation is accurate for points on the observed manifold when y i − y j 4 is sufficiently small with respect to the higher derivatives of the observation function.…”

“…For a more rigorous analysis, see [17]. Namely, the observed data follows a GMM distribution, with anisotropic covariance, induced by the Jacobian of the observation function.…”

Reconstruction of normal forms by learning informed observation geometries from data

“…However, as pointed out by Lafon 21, anisotropic diffusion maps can be computed by using different norms. This issue has been extensively studied recently, and several norms and metrics have been developed for this purpose (22)(23)(24).…”