Learning proper orthogonal decomposition of complex dynamics using heavy-ball neural ODEs
From MaRDI portal
Publication:6101554
Abstract: Proper orthogonal decomposition (POD) allows reduced-order modeling of complex dynamical systems at a substantial level, while maintaining a high degree of accuracy in modeling the underlying dynamical systems. Advances in machine learning algorithms enable learning POD-based dynamics from data and making accurate and fast predictions of dynamical systems. In this paper, we leverage the recently proposed heavy-ball neural ODEs (HBNODEs) [Xia et al. NeurIPS, 2021] for learning data-driven reduced-order models (ROMs) in the POD context, in particular, for learning dynamics of time-varying coefficients generated by the POD analysis on training snapshots generated from solving full order models. HBNODE enjoys several practical advantages for learning POD-based ROMs with theoretical guarantees, including 1) HBNODE can learn long-term dependencies effectively from sequential observations and 2) HBNODE is computationally efficient in both training and testing. We compare HBNODE with other popular ROMs on several complex dynamical systems, including the von K'{a}rm'{a}n Street flow, the Kurganov-Petrova-Popov equation, and the one-dimensional Euler equations for fluids modeling.
Recommendations
- Stochastic physics-informed neural ordinary differential equations
- NeuralPDE: modelling dynamical systems from data
- The neural network shifted-proper orthogonal decomposition: a machine learning approach for non-linear reduction of hyperbolic equations
- Deep learning nonlinear multiscale dynamic problems using Koopman operator
- Deep hidden physics models: deep learning of nonlinear partial differential equations
- Realization theory of recurrent neural ODEs using polynomial system embeddings
- FROM STANDARD APPROACH TO ODE NEURAL NETWORKS
- Solving flows of dynamical systems by deep neural networks and a novel deep learning algorithm
- POD-DL-ROM: enhancing deep learning-based reduced order models for nonlinear parametrized PDEs by proper orthogonal decomposition
Cites work
- A dynamic global-coefficient subgrid-scale eddy-viscosity model for large-eddy simulation in complex geometries
- A dynamic subgrid-scale eddy viscosity model
- A family of embedded Runge-Kutta formulae
- A Survey of Model Reduction by Balanced Truncation and Some New Results
- A survey of projection-based model reduction methods for parametric dynamical systems
- Absolute stability of global pattern formation and parallel memory storage by competitive neural networks
- Adaptive Semidiscrete Central-Upwind Schemes for Nonconvex Hyperbolic Conservation Laws
- An artificial neural network framework for reduced order modeling of transient flows
- Approximation of Large-Scale Dynamical Systems
- Construction of reduced-order models for fluid flows using deep feedforward neural networks
- Data-driven correction reduced order models for the quasi-geostrophic equations: a numerical investigation
- DIRECT NUMERICAL SIMULATION: A Tool in Turbulence Research
- Dynamic mode decomposition of numerical and experimental data
- How does momentum benefit deep neural networks architecture design? A few case studies
- scientific article; zbMATH DE number 3231758 (Why is no real title available?)
- Interpolatory model reduction of large-scale dynamical systems
- Machine learning closures for model order reduction of thermal fluids
- Neural network closures for nonlinear model order reduction
- Nonlinear analysis of hydrodynamic instability in laminar flames—I. Derivation of basic equations
- Nonlinear mode decomposition with convolutional neural networks for fluid dynamics
- On Flame Propagation Under Conditions of Stoichiometry
- On Upstream Differencing and Godunov-Type Schemes for Hyperbolic Conservation Laws
- Proper orthogonal decomposition and its applications-part. I: Theory
- Scheduled restart momentum for accelerated stochastic gradient descent
- Some methods of speeding up the convergence of iteration methods
Cited in
(7)- Time-series learning of latent-space dynamics for reduced-order model closure
- The neural network shifted-proper orthogonal decomposition: a machine learning approach for non-linear reduction of hyperbolic equations
- Neural-network learning of SPOD latent dynamics
- Linearly recurrent autoencoder networks for learning dynamics
- Machine learning enhanced Hankel dynamic-mode decomposition
- Regularized reduced order Lippmann-Schwinger-Lanczos method for inverse scattering problems in the frequency domain
- Reduced basis approximations of parameterized dynamical partial differential equations via neural networks
This page was built for publication: Learning proper orthogonal decomposition of complex dynamics using heavy-ball neural ODEs
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6101554)
