What problem does this paper attempt to address?

The paper primarily explores the variational form and numerical stability of Physics-Informed Neural Networks (PINNs) when dealing with forward and inverse problems in partial differential equations (PDEs). Specifically, the paper attempts to address the following key issues: 1. **Forward Problem**: Finding the solution \( u \) that satisfies the elliptic partial differential equation by minimizing the functional \( L_1 \), where \( u \) is the unknown minimizer, and \( f \) and \( g \) represent body force and surface force, respectively. The paper discusses the impact of the choice of the weight parameter \( \alpha \) on the problem and proves that the problem is well-defined in a specific weighted space \( V_\alpha \). 2. **Inverse Problem**: Simultaneously solving for the state \( u \) and the control variable \( f \) by minimizing another functional \( L_2 \), where \( u_d \) represents the observed values, and \( f_p \) encodes prior knowledge about the control variable. The paper particularly focuses on the case of partial observations and discusses the roles of the weight parameters \( \beta \) and \( \gamma \). It is proven that when \( 1/\gamma < \beta \) and the solution \( u \) satisfies \( H^2 \) regularity, the problem is also well-defined. 3. **Numerical Experiments**: The theoretical results are verified through numerical examples. First, the forward problem is discretized using Isogeometric Analysis (IGA), and the convergence rate in the \( H^2 \) space is validated. Then, the inverse problem is solved using both IGA and Neural Networks (NNs), comparing the performance differences between discretizations that satisfy the condition \( -\Delta V_{\beta,h} \subset G_{\beta,h} \) and those that do not. The results show that the discretization scheme satisfying this condition has better approximation effects. In summary, this paper aims to solve forward and inverse partial differential equation problems through variational methods and Physics-Informed Neural Networks, providing rigorous mathematical proofs and numerical validations.

Variational Formulations of the Strong Formulation -- Forward and Inverse Modeling using Isogeometric Analysis and Physics-Informed Networks

Effects of variational formulations on physics‐informed neural network performance in solid mechanics

Physics-informed graph neural Galerkin networks: A unified framework for solving PDE-governed forward and inverse problems

A Variational Formulation of Physics-Informed Neural Network for the Applications of Homogeneous and Heterogeneous Material Properties Identification

Variational Physics-Informed Neural Networks For Solving Partial Differential Equations

Physics-informed neural networks for transformed geometries and manifolds

A physics-informed deep learning framework for inversion and surrogate modeling in solid mechanics

Maximum-likelihood Estimators in Physics-Informed Neural Networks for High-dimensional Inverse Problems

Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations

Fully probabilistic deep models for forward and inverse problems in parametric PDEs

Robust Variational Physics-Informed Neural Networks

Inverse Physics-Informed Neural Networks for transport models in porous materials

Architectural Strategies for the optimization of Physics-Informed Neural Networks

Physics Informed Neural Networks for an Inverse Problem in Peridynamic Models

Enhanced physics‐informed neural networks for hyperelasticity

A mixed formulation for physics-informed neural networks as a potential solver for engineering problems in heterogeneous domains: Comparison with finite element method

Physics-informed neural networks for solving forward and inverse problems in complex beam systems

Physics-Informed neural network solver for numerical analysis in geoengineering

A Unified Framework for the Error Analysis of Physics-Informed Neural Networks

Solving Forward and Inverse Problems of Contact Mechanics using Physics-Informed Neural Networks