Discovery of dynamics using linear multistep methods

DOI10.1137/19M130981XMaRDI QIDQ5151929zbMATH OpenOpenAlexFDO

Authors Rachael T. Keller, Qiang Du

Publication date 23 February 2021

Published in SIAM Journal on Numerical Analysis (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1912.12728

zbMATH Keywords

artificial intelligencedata-driven modelingdiscovery of dynamicslearning dynamicslinear multistep methodsroot conditionstability and convergence

Mathematics Subject Classification ID

Multistep, Runge-Kutta and extrapolation methods for ordinary differential equations (65L06) Numerical solution of inverse problems involving ordinary differential equations (65L09) Stability and convergence of numerical methods for ordinary differential equations (65L20) Artificial intelligence (68T99)

Abstract: Linear multistep methods (LMMs) are popular time discretization techniques for the numerical solution of differential equations. Traditionally they are applied to solve for the state given the dynamics (the forward problem), but here we consider their application for learning the dynamics given the state (the inverse problem). This repurposing of LMMs is largely motivated by growing interest in data-driven modeling of dynamics, but the behavior and analysis of LMMs for discovery turn out to be significantly different from the well-known, existing theory for the forward problem. Assuming a highly idealized setting of being given the exact state with a zero residual of the discrete dynamics, we establish for the first time a rigorous framework based on refined notions of consistency and stability to yield convergence using LMMs for discovery. When applying these concepts to three popular $M-$step LMMs, the Adams-Bashforth, Adams-Moulton, and Backwards Differentiation Formula schemes, the new theory suggests that Adams-Bashforth for $M$ ranging from $1$ and $6$, Adams-Moulton for $M=0$ and $M=1$, and Backwards Differentiation Formula for all positive $M$ are convergent, and, otherwise, the methods are not convergent in general. In addition, we provide numerical experiments to both motivate and substantiate our theoretical analysis.

Recommendations

• The discovery of dynamics via linear multistep methods and deep learning: error estimation
• Data-driven sparse identification of nonlinear dynamical systems using linear multistep methods
• scientific article; zbMATH DE number 168447
• The stability problem for linear multistep methods: old and new results
• scientific article; zbMATH DE number 5353260

Cites work

• A data-driven approximation of the koopman operator: extending dynamic mode decomposition
• A kernel-based method for data-driven Koopman spectral analysis
• A note on unconditional fixed-h stability of linear multistep formulae
• A proposal on machine learning via dynamical systems
• A special stability problem for linear multistep methods
• A-BDF: A Generalization of the Backward Differentiation Formulae
• An Introduction to Numerical Analysis
• Bayesian deep convolutional encoder-decoder networks for surrogate modeling and uncertainty quantification
• Convergence and stability in the numerical integration of ordinary differential equations
• Data driven governing equations approximation using deep neural networks
• Data-driven discovery of closure models
• Data-driven science and engineering. Machine learning, dynamical systems, and control
• Deep hidden physics models: deep learning of nonlinear partial differential equations
• Deep learning
• Deep learning of dynamics and signal-noise decomposition with time-stepping constraints
• Discovering governing equations from data by sparse identification of nonlinear dynamical systems
• Foundations of machine learning
• scientific article; zbMATH DE number 3623496 (Why is no real title available?)
• scientific article; zbMATH DE number 1198839 (Why is no real title available?)
• scientific article; zbMATH DE number 1061412 (Why is no real title available?)
• scientific article; zbMATH DE number 940566 (Why is no real title available?)
• scientific article; zbMATH DE number 1418991 (Why is no real title available?)
• scientific article; zbMATH DE number 3182507 (Why is no real title available?)
• Lift \& learn: physics-informed machine learning for large-scale nonlinear dynamical systems
• Machine learning. A probabilistic perspective
• Machine learning: Trends, perspectives, and prospects
• Nonparametric inference of interaction laws in systems of agents from trajectory data
• Numerical Gaussian processes for time-dependent and nonlinear partial differential equations
• Numerical solution of ordinary differential equations.
• On the instability of high order backward-difference multistep methods
• On the Routh-Hurwitz Problem
• Pattern recognition and machine learning.
• PDE-Net 2.0: learning PDEs from data with a numeric-symbolic hybrid deep network
• Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• Prediction of Chaotic Time Series Based on the Recurrent Predictor Neural Network
• Solving high-dimensional partial differential equations using deep learning
• The stability properties ofq-step backward difference schemes
• Understanding machine learning. From theory to algorithms

Cited in

(22)

• Inferring switched nonlinear dynamical systems
• Weak SINDy for partial differential equations
• Discovery of subdiffusion problem with noisy data via deep learning
• VPNets: volume-preserving neural networks for learning source-free dynamics
• Kernel-based parameter estimation of dynamical systems with unknown observation functions
• The discovery of dynamics via linear multistep methods and deep learning: error estimation
• Weak SINDy: Galerkin-based data-driven model selection
• Data-driven sparse identification of nonlinear dynamical systems using linear multistep methods
• Deep Neural Networks for Solving Large Linear Systems Arising from High-Dimensional Problems
• A framework for machine learning of model error in dynamical systems
• Deep-OSG: deep learning of operators in semigroup
• On higher order drift and diffusion estimates for stochastic SINDy
• A pseudoreversible normalizing flow for stochastic dynamical systems with various initial distributions
• Error analysis based on inverse modified differential equations for discovery of dynamics using linear multistep methods and deep learning
• Implementation and (inverse modified) error analysis for implicitly templated ODE-nets
• DUE: a deep learning framework and library for modeling unknown equations
• High-order unconditionally energy-stable decoupled method for phase field modeling of pitting corrosion with adaptive implementation
• Ensemble adaptive libraries and inner-product sparse regression for robust nonlinear dynamics discovery
• Kernel Localization and Whole-Trajectory Generalization for Linear Multistep Methods in Deep Learning-Based Discovery of Dynamical Systems
• Approximation and error bounds for Kolmogorov-Arnold networks applications to dynamical system discovery via linear multistep methods
• Artifacts of numerical integration in learning dynamical systems
• Neural base-N encoding for solving massive nonlinear systems

Describes a project that uses

Uses Software

• PDE-Net
• AlexNet
• PRMLT
• PMTK
• RODAS
• ImageNet
• torchdiffeq

This page was built for publication: Discovery of dynamics using linear multistep methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5151929)