What problem does this paper attempt to address?

The paper primarily explores Physics-Informed Neural Networks (PINNs) and their applications and extensions in the field of scientific machine learning. Specifically: 1. **Problem Background**: Traditional numerical methods have achieved great success in solving partial differential equations (PDEs), but they still face challenges in seamlessly integrating data into existing algorithms, handling complex industrial problems, and solving inverse problems. Additionally, existing engineering software codes are large and difficult to maintain. 2. **Core Idea of PINNs**: By integrating data and mathematical models, PINNs define a new paradigm for scientific machine learning. This approach can handle data of different scales and combine partially known physical laws to infer unknown parameters or missing function terms, thereby simultaneously recovering solutions. 3. **Applications of PINNs**: PINNs can solve both forward and inverse problems and handle issues such as uncertainty quantification. Through automatic differentiation, PINNs can approximate the exact solutions of PDEs without using any regular grids. 4. **Extensions of PINNs**: - **Adaptive Weights**: A new gradient descent algorithm is proposed, which uses the Neural Tangent Kernel (NTK) to adaptively adjust the loss weights. - **Domain Decomposition**: Combining PINNs with domain decomposition methods to handle multi-scale problems, improving parallel speed in large-scale computational domains. - **Long-Time Integration**: By improving the loss function to explicitly consider physical causality, the problem of long-time dynamic system integration is addressed. - **Other Types of PDEs**: PINNs are also applied in fields such as stochastic partial differential equations and fractional partial differential equations. 5. **Theoretical Foundation**: Preliminary studies indicate that for certain types of PDEs, as the number of sampling points increases, PINNs converge to the true solution of the equations. In summary, this paper aims to demonstrate the potential and advantages of PINNs as a novel scientific machine learning tool in modeling complex physical systems and data analysis.

Physics-Informed Neural Networks and Extensions

Machine Learning Through Physics–Informed Neural Networks: Progress and Challenges

Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What's Next

Understanding Physics-Informed Neural Networks: Techniques, Applications, Trends, and Challenges

From PINNs to PIKANs: Recent Advances in Physics-Informed Machine Learning

Exploring Physics-Informed Neural Networks: From Fundamentals to Applications in Complex Systems

Physics-Informed Neural Network (PINN) Evolution and Beyond: A Systematic Literature Review and Bibliometric Analysis

fPINNs: Fractional Physics-Informed Neural Networks

Recipes for when physics fails: recovering robust learning of physics informed neural networks

An Expert's Guide to Training Physics-informed Neural Networks

A Review of Physics Informed Neural Networks for Multiscale Analysis and Inverse Problems

Physics-informed Neural Networks (PINN) for computational solid mechanics: Numerical frameworks and applications

Physics-Informed Neural Networks and Beyond: Enforcing Physical Constraints in Quantum Dissipative Dynamics

Physics-informed Neural Networks for Encoding Dynamics in Real Physical Systems

Investigations on Physics-Informed Neural Networks for Aerodynamics

SetPINNs: Set-based Physics-informed Neural Networks

Physics-informed attention-based neural network for solving non-linear partial differential equations

Computation of the Time-Dependent Dirac Equation with Physics-Informed Neural Networks

Architectural Strategies for the optimization of Physics-Informed Neural Networks

Numerical analysis of physics-informed neural networks and related models in physics-informed machine learning

Physics-informed Neural Networks with Unknown Measurement Noise