Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Recurrent Neural Networks for Partially Observed Dynamical Systems
(Submitted on 21 Sep 2021 (this version), latest version 5 Apr 2022 (v2))
Abstract: Complex nonlinear dynamics are ubiquitous in many fields. Moreover, we rarely have access to all of the relevant state variables governing the dynamics. Delay embedding allows us, in principle, to account for unobserved state variables. Here we provide an algebraic approach to delay embedding that permits explicit approximation of error. We also provide the asymptotic dependence of the first order approximation error on the system size. More importantly, this formulation of delay embedding can be directly implemented using a Recurrent Neural Network (RNN). This observation expands the interpretability of both delay embedding and RNN and facilitates principled incorporation of structure and other constraints into these approaches.
Submission history
From: Uttam Bhat [view email][v1] Tue, 21 Sep 2021 20:15:20 GMT (183kb,D)
[v2] Tue, 5 Apr 2022 01:36:40 GMT (157kb,D)
Link back to: arXiv, form interface, contact.