The Finite Neuron Method and Convergence Analysis

Abstract

We study a family of $H^m$-conforming piecewise polynomials based on artificial neural network, named as the finite neuron method (FNM), for numerical solution of $2m$-th order partial differential equations in ${\mathbb{R}}^d$ for any $m,d\ge 1$ and then provide convergence analysis for this method. Given a general domain $\Omega \subset {\mathbb{R}}^d$ and a partition ${\mathcal{T}}_h$ of $\Omega$, it is still an open problem in general how to construct conforming finite element subspace of $H^m(\Omega )$ that have adequate approximation properties. By using techniques from artificial neural networks, we construct a family of $H^m$-conforming set of functions consisting of piecewise polynomials of degree $k$ for any $k\ge m$ and we further obtain the error estimate when they are applied to solve elliptic boundary value problem of any order in any dimension. For example, the following error estimates between the exact solution $u$ and finite neuron approximation $u_N$ are obtained. $$\Vert u-u_N{\Vert }_{H^m(\Omega )}=\mathcal{O}(N^{-\frac{1}{2}-\frac{1}{d}}).$$ Discussions will also be given on the difference and relationship between the finite neuron method and finite element methods (FEM). For example, for finite neuron method, the underlying finite element grids are not given a priori and the discrete solution can only be obtained by solving a non-linear and non-convex optimization problem. Despite of many desirable theoretical properties of the finite neuron method analyzed in the paper, its practical value is a subject of further investigation since the aforementioned underlying non-linear and non-convex optimization problem can be expensive and challenging to solve. For completeness and also convenience to readers, some basic known results and their proofs are also included in this manuscript.

References in corpus (3)

• Multi-scale Deep Neural Network (MscaleDNN) for Solving Poisson-Boltzmann Equation in Complex Domains
• Representation formulas and pointwise properties for Barron functions
• A randomized Newton's method for solving differential equations based on the neural network discretization

Cited by in corpus (30)

• Local Extreme Learning Machines and Domain Decomposition for Solving Linear and Nonlinear Partial Differential Equations
• Sharp Bounds on the Approximation Rates, Metric Entropy, and $n$-widths of Shallow Neural Networks
• Convergence Rate Analysis for Deep Ritz Method
• A Modified Batch Intrinsic Plasticity Method for Pre-training the Random Coefficients of Extreme Learning Machines
• ReLU Deep Neural Networks from the Hierarchical Basis Perspective
• Near-Minimax Optimal Estimation With Shallow ReLU Neural Networks
• Approximation Properties of Deep ReLU CNNs
• Deep Ritz Method with Adaptive Quadrature for Linear Elasticity
• Imaging Conductivity from Current Density Magnitude using Neural Networks
• SVD-PINNs: Transfer Learning of Physics-Informed Neural Networks via Singular Value Decomposition
• Neural Control of Discrete Weak Formulations: Galerkin, Least-Squares and Minimal-Residual Methods with Quasi-Optimal Weights
• A Priori Analysis of Stable Neural Network Solutions to Numerical PDEs
• Characterization of the Variation Spaces Corresponding to Shallow Neural Networks
• Power Series Expansion Neural Network
• Runge-Kutta Physics Informed Neural Networks: Formulation and Analysis
• Spectral Barron space for deep neural network approximation
• A Hybrid Two-level MCMC Framework to Accelerate Posterior Mean Estimation with Deep Learning Surrogates for Bayesian Inverse Problems
• A neuron-wise subspace correction method for the finite neuron method
• Analysis of Deep Ritz Methods for Laplace Equations with Dirichlet Boundary Conditions
• Deep Ritz-Finite Element methods: Neural Network Methods trained with Finite Elements
• A Construction of $C^r$ Conforming Finite Element Spaces in Any Dimension
• Geometric decompositions of the simplicial lattice and smooth finite elements in arbitrary dimension
• A Natural Deep Ritz Method for Essential Boundary Value Problems
• Two neural-network-based methods for solving obstacle problems
• Embedding Inequalities for Barron-type Spaces
• Generalization Error Estimates of Machine Learning Methods for Solving High Dimensional Schrödinger Eigenvalue Problems
• Learn bifurcations of nonlinear parametric systems via equation-driven neural networks
• Two-level overlapping Schwarz preconditioners with universal coarse spaces for $2m$th-order elliptic problems
• Sharp Lower Bounds on the Approximation Rate of Shallow Neural Networks
• The regularity of electronic wave functions in Barron spaces