Learning Neural PDE Solvers with Convergence Guarantees

Abstract

Partial differential equations (PDEs) are widely used across the physical and computational sciences. Decades of research and engineering went into designing fast iterative solution methods. Existing solvers are general purpose, but may be sub-optimal for specific classes of problems. In contrast to existing hand-crafted solutions, we propose an approach to learn a fast iterative solver tailored to a specific domain. We achieve this goal by learning to modify the updates of an existing solver using a deep neural network. Crucially, our approach is proven to preserve strong correctness and convergence guarantees. After training on a single geometry, our model generalizes to a wide variety of geometries and boundary conditions, and achieves 2-3 times speedup compared to state-of-the-art solvers.

Cited by in corpus (24)

• Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers
• MgNet: A Unified Framework of Multigrid and Convolutional Neural Network
• FEA-Net: A Physics-guided Data-driven Model for Efficient Mechanical Response Prediction
• On some neural network architectures that can represent viscosity solutions of certain high dimensional Hamilton--Jacobi partial differential equations
• Enhancement of shock-capturing methods via machine learning
• Learning to Control PDEs with Differentiable Physics
• Variational training of neural network approximations of solution maps for physical models
• FiniteNet: A Fully Convolutional LSTM Network Architecture for Time-Dependent Partial Differential Equations
• Belief Propagation Neural Networks
• LordNet: An Efficient Neural Network for Learning to Solve Parametric Partial Differential Equations without Simulated Data
• Reinforcement Learning for Adaptive Mesh Refinement
• Overcoming the curse of dimensionality for some Hamilton--Jacobi partial differential equations via neural network architectures
• Frequency-compensated PINNs for Fluid-dynamic Design Problems
• Learning optimal multigrid smoothers via neural networks
• Optimization-Based Algebraic Multigrid Coarsening Using Reinforcement Learning
• Analysis of Deep Ritz Methods for Laplace Equations with Dirichlet Boundary Conditions
• Implicit Neural Solver for Time-dependent Linear PDEs with Convergence Guarantee
• Neural Multigrid Architectures
• Constrained Linear Data-feature Mapping for Image Classification
• Learning invariance preserving moment closure model for Boltzmann-BGK equation
• Semi-Implicit Neural Solver for Time-dependent Partial Differential Equations
• Learning To Solve Differential Equations Across Initial Conditions
• Physics-Augmented Learning: A New Paradigm Beyond Physics-Informed Learning
• SEMULATOR: Emulating the Dynamics of Crossbar Array-based Analog Neural System with Regression Neural Networks