Physics-Informed Neural Networks (PINN) Testing

Introduction

A playground repo for testing some Physics-informed Neural Networks (PINNs) for solving simple differential equations. The toy models will be based on the formulation of M. Raissi, P. Perdikaris and G.E. Karniadakis and are meant as an initial exploration of this new topic from personal interest.

Models

• Solving_ODE.ipynb: A toy model solution to first and second order ODEs (radioactive decay and a harmonic oscillator) for demonstration purposes

Resources

Given the limited number of formal resources online, I used the following videos to get a better idea of the hands-on implementation details:

• Matthieu Barreau - Physics-Informed Learning: Using Neural Networks to Solve Differential Equations
• A Hands-on Introduction to Physics-informed Machine Learning
• Get hands On with PINNs, which uses the library DeepXDE
• to be continued