Abstract:This paper studies the shallow Ritz method for solving the one-dimensional diffusion problem. It is shown that the shallow Ritz method improves the order of approximation dramatically for non-smooth problems. To realize this optimal or nearly optimal order of the shallow Ritz approximation, we develop a damped block Newton (dBN) method that alternates between updates of the linear and non-linear parameters. Per each iteration, the linear and the non-linear parameters are updated by exact inversion and one step of a modified, damped Newton method applied to a reduced non-linear system, respectively. The computational cost of each dBN iteration is $O(n)$. Starting with the non-linear parameters as a uniform partition of the interval, numerical experiments show that the dBN is capable of efficiently moving mesh points to nearly optimal locations. To improve efficiency of the dBN further, we propose an adaptive damped block Newton (AdBN) method by combining the dBN with the adaptive neuron enhancement (ANE) method [26].

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to effectively improve the approximation order of the shallow - Ritz method in one - dimensional diffusion problems, especially for non - smooth problems.** Specifically, the paper focuses on how to achieve the optimal or near - optimal approximation order by optimizing linear and nonlinear parameters when using the shallow - Ritz method to solve one - dimensional diffusion problems. To achieve this goal, the author developed a damped block Newton (dBN) method, which alternately updates linear and nonlinear parameters, and the computational cost of each iteration step is $O(n)$. In addition, in order to further improve the efficiency of the dBN method, the author proposed the Adaptive Damped Block Newton (AdBN) method, which combines dBN and the Adaptive Neuron Enhancement (ANE) method. AdBN can not only quickly and effectively move grid points to near - optimal positions, but also achieve a near - optimal approximation order in some cases. ### Key Problem Summary: 1. **Non - convex Optimization Problem**: Determining the optimal configuration of nonlinear parameters (i.e., grid point positions) in a shallow ReLU neural network is a complex non - convex optimization problem. 2. **Approximation Order Improvement**: For non - smooth problems, the shallow - Ritz method can significantly improve the approximation order, but an effective algorithm is required to optimize the parameters. 3. **Computational Efficiency**: Traditional optimization methods such as BFGS are less efficient in handling such problems, so more efficient algorithms need to be developed. ### Solutions: - **dBN Method**: Solve the nonlinear system by alternately updating linear and nonlinear parameters, using exact inversion and the damped Newton method. - **AdBN Method**: Combine dBN and ANE methods and dynamically add new neurons to improve the approximation effect. ### Main Contributions: - Proposed the dBN and AdBN methods, which solve the difficult problem of nonlinear parameter optimization in the shallow - Ritz method. - Proved the high efficiency and accuracy of these methods in numerical experiments, especially their performance in handling non - smooth problems is better than that of traditional methods. Hope this summary can help you understand the core problems and solutions of the paper. If you have more specific questions or need further explanation, please feel free to let me know!

Efficient Shallow Ritz Method For 1D Diffusion Problems

High Accuracy Non-Equidistant Method for Singular Perturbation Reaction-Diffusion Problem

Fast Iterative Solver For Neural Network Method: II. 1D Diffusion-Reaction Problems And Data Fitting

Efficient L1-ADI finite difference method for the two-dimensional nonlinear time-fractional diffusion equation

Efficient Numerical Algorithms for Three-Dimensional Fractional Partial Differential Equations

Efficient second-order ADI difference schemes for three-dimensional Riesz space-fractional diffusion equations

Optimal finite-differences discretization for the diffusion equation from the perspective of large-deviation theory

A rational high-order compact ADI method for unsteady convection–diffusion equations

Two fast and unconditionally stable finite difference methods for Riesz fractional diffusion equations with variable coefficients

A reduced-order model for advection-dominated problems based on Radon-Cumulative-Distribution Transform

Deep Ritz Method with Adaptive Quadrature for Linear Elasticity

Analysis and efficient implementation of alternating direction implicit finite volume method for Riesz space‐fractional diffusion equations in two space dimensions

A novel meshless method based on RBF for solving variable-order time fractional advection-diffusion-reaction equation in linear or nonlinear systems

An Efficient Adaptive Mesh Redistribution Method for a Non-Linear Dirac Equation

An Implicit Scheme of the 1D Convection-Diffusion-Reaction Equation

Numerical Methods for Linear Diffusion Equations in the Presence of an Interface

A novel fourth-order scheme for two-dimensional Riesz space fractional nonlinear reaction-diffusion equations and its optimal preconditioned solver

Deep Ritz Methods for Laplace Equations with Dirichlet Boundary Condition

A shallow Ritz method for elliptic problems with singular sources

An Implicit Numerical Method for the Riemann–Liouville Distributed-Order Space Fractional Diffusion Equation

Fast Iterative Method with a Second-Order Implicit Difference Scheme for Time-Space Fractional Convection–Diffusion Equation