Physics-Informed Echo State Networks for Chaotic Systems Forecasting - Archive ouverte HAL Access content directly
Book Sections Year : 2019

Physics-Informed Echo State Networks for Chaotic Systems Forecasting

Abstract

We propose a physics-informed Echo State Network (ESN) to predict the evolution of chaotic systems. Compared to conventional ESNs, the physics-informed ESNs are trained to solve supervised learning tasks while ensuring that their predictions do not violate physical laws. This is achieved by introducing an additional loss function during the training of the ESNs, which penalizes non-physical predictions without the need of any additional training data. This approach is demonstrated on a chaotic Lorenz system, where the physics-informed ESNs improve the predictability horizon by about two Lyapunov times as compared to conventional ESNs. The proposed framework shows the potential of using machine learning combined with prior physical knowledge to improve the time-accurate prediction of chaotic dynamical systems.
Fichier principal
Vignette du fichier
PBNN.pdf (362.26 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02111699 , version 1 (26-04-2019)

Identifiers

Cite

Nguyen Anh Khoa Doan, Wolfgang Polifke, Luca Magri. Physics-Informed Echo State Networks for Chaotic Systems Forecasting. Lecture Notes in Computer Science book series (LNCS), 11539, pp.192-198, 2019, ⟨10.1007/978-3-030-22747-0_15⟩. ⟨hal-02111699⟩

Collections

TDS-MACS
210 View
244 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More