A Neural Network Approach for Homogenization of Multiscale Problems
From MaRDI portal
Abstract: We propose a neural network-based approach to the homogenization of multiscale problems. The proposed method uses a derivative-free formulation of a training loss, which incorporates Brownian walkers to find the macroscopic description of a multiscale PDE solution. Compared with other network-based approaches for multiscale problems, the proposed method is free from the design of hand-crafted neural network architecture and the cell problem to calculate the homogenization coefficient. The exploration neighborhood of the Brownian walkers affects the overall learning trajectory. We determine the bounds of micro- and macro-time steps that capture the local heterogeneous and global homogeneous solution behaviors, respectively, through a neural network. The bounds imply that the computational cost of the proposed method is independent of the microscale periodic structure for the standard periodic problems. We validate the efficiency and robustness of the proposed method through a suite of linear and nonlinear multiscale problems with periodic and random field coefficients.
Recommendations
- Neural Network Approximation of Coarse-Scale Surrogates in Numerical Homogenization
- Physics-informed neural networks for learning the homogenized coefficients of multiscale elliptic equations
- A neural network multigrid solver for the Navier-Stokes equations
- Computational homogenization of nonlinear elastic materials using neural networks
- Neural‐network‐based approximations for solving partial differential equations
- Application of multigrid iterative neural network method to stationary problems of advection-diffusion
- Approximation de calculs éléments finis par réseaux neuronaux: quelques applications en optimisation
- Neural network method for solving partial differential equations
- NH-PINN: neural homogenization-based physics-informed neural network for multiscale problems
- An Efficient Neural-Network and Finite-Difference Hybrid Method for Elliptic Interface Problems with Applications
Cites work
- A derivative-free method for solving elliptic partial differential equations with deep neural networks
- Adaptive activation functions accelerate convergence in deep and physics-informed neural networks
- Asymptotic analysis for periodic structures
- Asymptotic preserving (AP) schemes for multiscale kinetic and hyperbolic equations: a review
- Averaging methods in nonlinear dynamical systems
- Deep reinforcement learning of viscous incompressible flow
- DGM: a deep learning algorithm for solving partial differential equations
- Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networks
- Generalized multiscale finite element methods (GMsFEM)
- Heterogeneous multiscale methods for stiff ordinary differential equations
- scientific article; zbMATH DE number 4174505 (Why is no real title available?)
- scientific article; zbMATH DE number 51724 (Why is no real title available?)
- LARGE-EDDY SIMULATION OF TURBULENT COMBUSTION
- Multiscale numerical methods for passive advection-diffusion in incompressible turbulent flow fields
- NH-PINN: neural homogenization-based physics-informed neural network for multiscale problems
- On the eigenvector bias of Fourier feature networks: from regression to solving multi-scale PDEs with physics-informed neural networks
- Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
- Solving high-dimensional partial differential equations using deep learning
- The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems
- The heterogeneous multiscale method
- The mean distance to the n th neighbour in a uniform distribution of random points: an application of probability theory
- Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks
- Variable step size multiscale methods for stiff and highly oscillatory dynamical systems
- When and why PINNs fail to train: a neural tangent kernel perspective
Cited in
(18)- CENN: conservative energy method based on neural networks with subdomains for solving variational problems involving heterogeneous and complex geometries
- NH-PINN: neural homogenization-based physics-informed neural network for multiscale problems
- Solving multiscale steady radiative transfer equation using neural networks with uniform stability
- Physics-informed neural networks for learning the homogenized coefficients of multiscale elliptic equations
- Neural network-based variational methods for solving quadratic porous medium equations in high dimensions
- scientific article; zbMATH DE number 5866702 (Why is no real title available?)
- A Multiscale Neural Network Based on Hierarchical Matrices
- Neural Network Approximation of Coarse-Scale Surrogates in Numerical Homogenization
- Concurrent multiscale simulations of nonlinear random materials using probabilistic learning
- Bayesian deep operator learning for homogenized to fine-scale maps for multiscale PDE
- Learning homogenization for elliptic operators
- A stochastic approach for elliptic problems in perforated domains
- A hybrid physics-informed neural network based multiscale solver as a partial differential equation constrained optimization problem
- Singular layer physics-informed neural network method for convection-dominated boundary layer problems in two dimensions
- Multiscale data-driven modeling for nonlinear dynamical system with external inputs in porous media
- An analysis of the derivative-free loss method for solving PDEs
- A derivative-free framework for capturing macroscopic behavior of incompressible turbulent flows
- A stochastic vanishing viscosity approach for eikonal equations
This page was built for publication: A Neural Network Approach for Homogenization of Multiscale Problems
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6178099)
