Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks

We study the use of feedforward neural networks (FNN) to develop models of nonlinear dynamical systems from data. Emphasis is placed on predictions at long times, with limited data availability. Inspired by global stability analysis, and the observation of strong correlation between the local error...

Full description

Saved in:
Bibliographic Details
Main Authors: Shaowu Pan, Karthik Duraisamy
Format: Article
Language:English
Published: Wiley 2018-01-01
Series:Complexity
Online Access:http://dx.doi.org/10.1155/2018/4801012
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850173031852277760
author Shaowu Pan
Karthik Duraisamy
author_facet Shaowu Pan
Karthik Duraisamy
author_sort Shaowu Pan
collection DOAJ
description We study the use of feedforward neural networks (FNN) to develop models of nonlinear dynamical systems from data. Emphasis is placed on predictions at long times, with limited data availability. Inspired by global stability analysis, and the observation of strong correlation between the local error and the maximal singular value of the Jacobian of the ANN, we introduce Jacobian regularization in the loss function. This regularization suppresses the sensitivity of the prediction to the local error and is shown to improve accuracy and robustness. Comparison between the proposed approach and sparse polynomial regression is presented in numerical examples ranging from simple ODE systems to nonlinear PDE systems including vortex shedding behind a cylinder and instability-driven buoyant mixing flow. Furthermore, limitations of feedforward neural networks are highlighted, especially when the training data does not include a low dimensional attractor. Strategies of data augmentation are presented as remedies to address these issues to a certain extent.
format Article
id doaj-art-cbdb8edcae8340cda6ef0e84e7058721
institution OA Journals
issn 1076-2787
1099-0526
language English
publishDate 2018-01-01
publisher Wiley
record_format Article
series Complexity
spelling doaj-art-cbdb8edcae8340cda6ef0e84e70587212025-08-20T02:19:57ZengWileyComplexity1076-27871099-05262018-01-01201810.1155/2018/48010124801012Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural NetworksShaowu Pan0Karthik Duraisamy1Department of Aerospace Engineering, University of Michigan, Ann Arbor, MI, USADepartment of Aerospace Engineering, University of Michigan, Ann Arbor, MI, USAWe study the use of feedforward neural networks (FNN) to develop models of nonlinear dynamical systems from data. Emphasis is placed on predictions at long times, with limited data availability. Inspired by global stability analysis, and the observation of strong correlation between the local error and the maximal singular value of the Jacobian of the ANN, we introduce Jacobian regularization in the loss function. This regularization suppresses the sensitivity of the prediction to the local error and is shown to improve accuracy and robustness. Comparison between the proposed approach and sparse polynomial regression is presented in numerical examples ranging from simple ODE systems to nonlinear PDE systems including vortex shedding behind a cylinder and instability-driven buoyant mixing flow. Furthermore, limitations of feedforward neural networks are highlighted, especially when the training data does not include a low dimensional attractor. Strategies of data augmentation are presented as remedies to address these issues to a certain extent.http://dx.doi.org/10.1155/2018/4801012
spellingShingle Shaowu Pan
Karthik Duraisamy
Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks
Complexity
title Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks
title_full Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks
title_fullStr Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks
title_full_unstemmed Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks
title_short Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks
title_sort long time predictive modeling of nonlinear dynamical systems using neural networks
url http://dx.doi.org/10.1155/2018/4801012
work_keys_str_mv AT shaowupan longtimepredictivemodelingofnonlineardynamicalsystemsusingneuralnetworks
AT karthikduraisamy longtimepredictivemodelingofnonlineardynamicalsystemsusingneuralnetworks