Abstract:The research in Artificial Intelligence methods with potential applications in science has become an essential task in the scientific community last years. Physics Informed Neural Networks (PINNs) is one of this methods and represent a contemporary technique that is based on the fundamentals of neural networks to solve differential equations. These kind of networks have the potential to improve or complement classical numerical methods in computational physics, making them an exciting area of study. In this paper, we introduce PINNs at an elementary level, mainly oriented to physics education so making them suitable for educational purposes at both undergraduate and graduate levels. PINNs can be used to create virtual simulations and educational tools that aid in understating complex physical concepts and processes where differential equations are involved. By combining the power of neural networks with physics principles, PINNs can provide an interactive and engaging learning experience that can improve students' understanding and retention of physics concepts in higher education.

What problem does this paper attempt to address?

The main objective of this paper is to introduce how to use Physics-Informed Neural Networks (PINNs) to solve differential equations and to demonstrate their application in education through several classic examples of ordinary differential equations (ODEs). Specifically: 1. **Introduction to the Basic Principles of PINNs**: The paper first introduces the basic knowledge of neural networks and how to calculate gradients through Automatic Differentiation and Backpropagation algorithms to minimize the loss function. 2. **Application Examples**: - **First-order Ordinary Differential Equations (First-order ODEs)**: Using the exponential decay equation as an example, it demonstrates how to construct PINNs to solve such equations and verifies the model's interpolation and extrapolation capabilities within the training interval. - **Second-order Ordinary Differential Equations (Second-order ODEs)**: Through the harmonic oscillator equation (describing spring-mass systems, small amplitude pendulums, etc.), it illustrates the application of PINNs and discusses the oscillation phenomena that may occur during the training process and their causes. - **Nonlinear Second-order Ordinary Differential Equations (Nonlinear Second-order ODEs)**: Using the Korteweg-de Vries equation as an example, it explores the ability of PINNs to handle nonlinear problems, especially focusing on special phenomena such as solitons. 3. **Educational Value**: By combining the power of neural networks with physical principles, PINNs can provide an interactive and engaging learning experience, helping to enhance students' understanding and memory of complex physical concepts, with broad application prospects in higher education. In summary, this paper aims to introduce beginners to how to use PINNs to solve various types of differential equations through specific mathematical examples and Python code implementations, and emphasizes the potential value of this method in the field of education.

Solving differential equations with Deep Learning: a beginner's guide

PinnDE: Physics-Informed Neural Networks for Solving Differential Equations

Solving differential equations using physics informed deep learning: a hand-on tutorial with benchmark tests

Solving Partial Differential Equations Using Deep Learning and Physical Constraints

A Survey on Solving and Discovering Differential Equations Using Deep Neural Networks

DeepXDE: A Deep Learning Library for Solving Differential Equations

Solving Partial Differential Equations with Point Source Based on Physics-Informed Neural Networks

A hands-on introduction to Physics-Informed Neural Networks for solving partial differential equations with benchmark tests taken from astrophysics and plasma physics

Exploration on Physics-Informed Neural Networks on Partial Differential Equations (Student Abstract)

The Old and the New: Can Physics-Informed Deep-Learning Replace Traditional Linear Solvers?

One-Shot Transfer Learning of Physics-Informed Neural Networks

Artificial Neural Networks Solutions for Solving Differential Equations: A Focus and Example for Flow of Viscoelastic Fluid with Microrotation

On Physics-Informed Deep Learning for Solving Navier-Stokes Equations

Solving stiff ordinary differential equations using physics informed neural networks (PINNs): simple recipes to improve training of vanilla-PINNs

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

A General Method for Solving Differential Equations of Motion Using Physics-Informed Neural Networks

A physics-informed deep learning approach for solving strongly degenerate parabolic problems

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

Self-Paced Learning Enhanced Physics-informed Neural Networks for Solving Partial Differential Equations

ST-PINN: A Self-Training Physics-Informed Neural Network for Partial Differential Equations

A Deep Learning Framework for Solving Hyperbolic Partial Differential Equations: Part I