Failure-Informed Adaptive Sampling for PINNs
From MaRDI portal
Publication:6175124
Abstract: Physics-informed neural networks (PINNs) have emerged as an effective technique for solving PDEs in a wide range of domains. It is noticed, however, the performance of PINNs can vary dramatically with different sampling procedures. For instance, a fixed set of (prior chosen) training points may fail to capture the effective solution region (especially for problems with singularities). To overcome this issue, we present in this work an adaptive strategy, termed the failure-informed PINNs (FI-PINNs), which is inspired by the viewpoint of reliability analysis. The key idea is to define an effective failure probability based on the residual, and then, with the aim of placing more samples in the failure region, the FI-PINNs employs a failure-informed enrichment technique to adaptively add new collocation points to the training set, such that the numerical accuracy is dramatically improved. In short, similar as adaptive finite element methods, the proposed FI-PINNs adopts the failure probability as the posterior error indicator to generate new training points. We prove rigorous error bounds of FI-PINNs and illustrate its performance through several problems.
Recommendations
- Probabilistic analysis of adaptative sampling
- scientific article; zbMATH DE number 7088254
- scientific article; zbMATH DE number 805117
- On adaptive sampling
- Adaptive sampling schemes for density estimation
- Adaptive sampling of large deviations
- A prediction approach in adaptive sampling
- Analysis and application of adaptive sampling
- Adaptive sampling strategies for stochastic optimization
Cites work
- A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks
- An Augmented Lagrangian Deep Learning Method for Variational Problems with Essential Boundary Conditions
- DeepXDE: a deep learning library for solving differential equations
- DGM: a deep learning algorithm for solving partial differential equations
- Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations
- Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems
- Machine learning and computational mathematics
- Monte Carlo fPINNs: deep learning method for forward and inverse problems involving high dimensional fractional partial differential equations
- Normalizing field flows: solving forward and inverse stochastic differential equations using physics-informed flow models
- Pattern recognition and machine learning.
- Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
- Self-adaptive physics-informed neural networks
- Solving time dependent Fokker-Planck equations via temporal normalizing flow
- The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems
- Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks
- Weak adversarial networks for high-dimensional partial differential equations
- When and why PINNs fail to train: a neural tangent kernel perspective
Cited in
(78)- HRW: Hybrid Residual and Weak Form Loss for Solving Elliptic Interface Problems with Neural Network
- PI-VEGAN: Physics Informed Variational Embedding Generative Adversarial Networks for Stochastic Differential Equations
- Improved training of physics-informed neural networks for parabolic differential equations with sharply perturbed initial conditions
- Feature-adjacent multi-fidelity physics-informed machine learning for partial differential equations
- Less Emphasis on Hard Regions: Curriculum Learning of PINNs for Singularly Perturbed Convection-Diffusion-Reaction Problems
- Correcting model misspecification in physics-informed neural networks (PINNs)
- Render unto numerics: orthogonal polynomial neural operator for PDEs with nonperiodic boundary conditions
- Solving Poisson problems in polygonal domains with singularity enriched physics informed neural networks
- Domain decomposition learning methods for solving elliptic problems
- Plasma-simulation physics informed neural networks (PS-PINNs) for global discharge models
- Failure-informed adaptive sampling for PINNs. II: Combining with re-sampling and subset simulation
- Deep finite volume method for partial differential equations
- An accurate and efficient continuity-preserved method based on randomized neural networks for elliptic interface problems
- A new method to compute the blood flow equations using the physics-informed neural operator
- Deep adaptive sampling for surrogate modeling without labeled data
- MHDnet: physics-preserving learning for solving magnetohydrodynamics problems
- A neural particle method with interface tracking and adaptive particle refinement for free surface flows
- Loss jump during loss switch in solving PDEs with neural networks
- Efficiently training physics-informed neural networks via anomaly-aware optimization
- Deep mixed residual method for solving PDE-constrained optimization problems
- Adaptive operator learning for infinite-dimensional Bayesian inverse problems
- Time integration schemes based on neural networks for solving partial differential equations on coarse grids
- Annealed adaptive importance sampling method in PINNs for solving high dimensional partial differential equations
- An asymptotic expansion-based deep neural network for solving singularly perturbed problems with exponential boundary layers
- The deep Ritz methods for the elastic contact problems with the Signorini, Tresca-friction, Coulomb-friction, and non-convex non-smooth subgradient-type boundary conditions
- Energy-based adaptive deep unfitted Nitsche method for elliptic interface problems
- TSPINN: Thompson sampling-based adaptive training for physics-informed neural networks
- Adaptive pseudo-solution-assisted PINNs for multiscale Darcy and Darcy-Forchheimer equations
- Variational based NN and PINN for the Poisson problem
- Adaptive neural network basis methods for partial differential equations with low-regular solutions
- Multi-benchmark adaptive sampling physics-informed neural network for complex and coupled equations
- Deep collocation method: a framework for solving PDEs using neural networks with error control
- Estimating committor functions via deep adaptive sampling on rare transition paths
- Self-adaptive weights based on balanced residual decay rate for physics-informed neural networks and deep operator networks
- High-order multi-scale neural network method for quasi-static thermo-mechanical problems of composite materials
- Scalable Bayesian physics-informed Kolmogorov-Arnold networks
- Two-scale neural networks for partial differential equations with small parameters
- Failure-informed adaptive sampling for PINNs. III: Applications to inverse problems
- A coordinate transformation-based physics-informed neural networks for hyperbolic conservation laws
- Adaptive importance sampling for deep Ritz
- A hybrid iterative method for elliptic variational inequalities of the second kind
- Data generation-based operator learning for solving partial differential equations on unbounded domains
- An imbalanced learning-based sampling method for physics-informed neural networks
- SOC-MartNet: a martingale neural network for the Hamilton-Jacobi-Bellman equation without explicit \(\inf_{u\in U}H\) in stochastic optimal controls
- An adaptive optimal selection approach of the mixture-of-experts model embedded with PINNs for one-dimensional hyperbolic conservation laws
- Regularity-conforming neural networks (ReCoNNs) for solving partial differential equations
- Initialization-enhanced physics-informed neural network with domain decomposition (IDPINN)
- On understanding and overcoming spectral biases of deep neural network learning methods for solving PDEs
- Accurate adaptive deep learning method for solving elliptic problems
- Adaptive interface-PINNs (AdaI-PINNs): an efficient physics-informed neural networks framework for interface problems
- Multi-objective loss balancing for physics-informed deep learning
- Gradient flow based phase-field modeling using separable neural networks
- Adaptive trajectories sampling for solving PDEs with deep learning methods
- Mitigating distribution shift in machine learning-augmented hybrid simulation
- IB-UQ: information bottleneck based uncertainty quantification for neural function regression and neural operator learning
- Frequency-adaptive multi-scale deep neural networks
- Continuity-preserved deep learning method for solving elliptic interface problems
- Resolution invariant deep operator network for PDEs with complex geometries
- Stacked networks improve physics-informed training: applications to neural networks and deep operator networks
- Noise robust physics-informed generative adversarial networks for solving stochastic differential equations
- R-PINN: recovery-type a-posteriori estimator enhanced adaptive PINN
- Complex physics-informed neural network
- An adaptive error estimator for stationary Navier-Stokes equations using variational physics informed neural networks
- Deep random difference method for high-dimensional quasilinear parabolic partial differential equations
- Convolution-weighting method for the physics-informed neural network: a primal-dual optimization perspective
- Kernel-based PINNs for time fractional differential equations and Volterra integral-differential equations
- Stochastic quadrature rules for solving PDEs using neural networks
- Domain decomposition methods with Physics-informed neural networks for elliptic equations on manifolds
- Residual-Christoffel Sampling for Random Feature Collocation of Linear PDEs
- A Hybrid Physics-Informed Neural Network Framework for Subcritical and Supercritical Dynamics in Multi-Species Chemotaxis
- Shallow neural networks for solving nonuniform size-modified Poisson-Boltzmann equations
- Consensus-based adaptive sampling and approximation for high-dimensional energy landscapes
- A structure-informed framework for solving time-dependent partial differential equations with neural networks
- A comparative study of neural network solvers for second-order boundary value problems
- Neural networks for solving elliptic problems: analysis and adaptivities
- The adaptive solution of high-frequency Helmholtz equations via multi-grade deep learning
- Order statistical adaptive residual loss for physics-informed neural networks
- MCMC-PINNs: adaptively sampling collocation points of PINNs with an improved Markov chain Monte Carlo method
This page was built for publication: Failure-Informed Adaptive Sampling for PINNs
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6175124)
