Neural Galerkin Schemes with Active Learning for High-Dimensional Evolution Equations
arXiv:2203.01360 · doi:10.1016/j.jcp.2023.112588
Abstract
Deep neural networks have been shown to provide accurate function approximations in high dimensions. However, fitting network parameters requires informative training data that are often challenging to collect in science and engineering applications. This work proposes Neural Galerkin schemes based on deep learning that generate training data with active learning for numerically solving high-dimensional partial differential equations. Neural Galerkin schemes build on the Dirac-Frenkel variational principle to train networks by minimizing the residual sequentially over time, which enables adaptively collecting new training data in a self-informed manner that is guided by the dynamics described by the partial differential equations. This is in contrast to other machine learning methods that aim to fit network parameters globally in time without taking into account training data acquisition. Our finding is that the active form of gathering training data of the proposed Neural Galerkin schemes is key for numerically realizing the expressive power of networks in high dimensions. Numerical experiments demonstrate that Neural Galerkin schemes have the potential to enable simulating phenomena and processes with many variables for which traditional and other deep-learning-based solvers fail, especially when features of the solutions evolve locally such as in high-dimensional wave propagation problems and interacting particle systems described by Fokker-Planck and kinetic equations.
References in corpus (18)
- DGM: A deep learning algorithm for solving partial differential equations
- Solving high-dimensional partial differential equations using deep learning
- Machine learning accelerated computational fluid dynamics
- Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations
- A unified deep artificial neural network approach to partial differential equations in complex geometries
- Learning data driven discretizations for partial differential equations
- Lift & Learn: Physics-informed machine learning for large-scale nonlinear dynamical systems
- A fast and accurate physics-informed neural network reduced order model with shallow masked autoencoder
- Deep learning of dynamics and signal-noise decomposition with time-stepping constraints
- Computing Committor Functions for the Study of Rare Events Using Deep Learning
- Evolutional Deep Neural Network
- Nonlinear model reduction on metric spaces. Application to one-dimensional conservative PDEs in Wasserstein spaces
- Projection-Based Model Reduction with Dynamically Transformed Modes
- Neural Implicit Flow: a mesh-agnostic dimensionality reduction paradigm of spatio-temporal data
- Dynamical model reduction method for solving parameter-dependent dynamical systems
- Derivative-Informed Projected Neural Networks for High-Dimensional Parametric Maps Governed by PDEs
- Some observations on high-dimensional partial differential equations with Barron data
- HyperPINN: Learning parameterized differential equations with physics-informed hypernetworks