ETH Zürich DLSC: Physics-Informed Neural Networks - Applications

↓↓↓ LECTURE OVERVIEW BELOW ↓↓↓
ETH Zürich Deep Learning in Scientific Computing 2023
Lecture 5: Physics-Informed Neural Networks - Applications
Course Website (links to slides and tutorials): camlab.ethz.ch/teaching/deep-...
Lecturers: Ben Moseley and Siddhartha Mishra
▬ Lecture Content ▬▬▬▬▬▬▬▬▬
0:00 - Lecture overview
1:45 - What is a physics-informed neural network (PINN)?
11:59 - PINNs as a general framework
17:17 - PINNs for solving the Burgers' equation
20:20 - How to train PINNs
28:34 - 🔴 Live coding a PINN - part 1 | Code: github.com/benmoseley/DLSC-2023
39:42 - Training considerations
44:27 - [break - please skip]
53:07 - Simulation with PINNs
1:00:14 - Solving inverse problems with PINNs
1:14:00 - 🔴 Live coding a PINN - part 2 | Code: github.com/benmoseley/DLSC-2023
1:24:10 - Equation discovery with PINNs
▬ Course Overview ▬▬▬▬▬▬▬▬▬
Lecture 1: Course Introduction • ETH Zürich DLSC: Cours...
Lecture 2: Introduction to Deep Learning Part 1 • ETH Zürich DLSC: Intro...
Lecture 3: Introduction to Deep Learning Part 2 • ETH Zürich DLSC: Intro...
Lecture 4: Physics-Informed Neural Networks - Introduction • ETH Zürich DLSC: Physi...
Lecture 5: Physics-Informed Neural Networks - Applications • ETH Zürich DLSC: Physi...
Lecture 6: Physics-Informed Neural Networks - Limitations and Extensions • ETH Zürich DLSC: Physi...
Lecture 7: Introduction to Operator Learning Part 1 • ETH Zürich DLSC: Intro...
Lecture 8: Introduction to Operator Learning Part 2 • ETH Zürich DLSC: Intro...
Lecture 9: Deep Operator Networks • ETH Zürich DLSC: Deep ...
Lecture 10: Neural Operators • ETH Zürich DLSC: Neura...
Lecture 11: Fourier Neural Operators and Convolutional Neural Operators • ETH Zürich DLSC: Fouri...
Lecture 12: Introduction to Differentiable Physics Part 1 • ETH Zürich DLSC: Intro...
Lecture 13: Introduction to Differentiable Physics Part 2 • ETH Zürich DLSC: Intro...
▬ Course Learning Objectives ▬▬▬▬▬
The objective of this course is to introduce students to advanced applications of deep learning in scientific computing. The focus will be on the design and implementation of algorithms as well as on the underlying theory that guarantees reliability of the algorithms. We provide several examples of applications in science and engineering where deep learning based algorithms outperform state of the art methods.
By the end of the course you should be:
- Aware of advanced applications of deep learning in scientific computing
- Familiar with the design, implementation and theory of these algorithms
- Understand the pros/cons of using deep learning
- Understand key scientific machine learning concepts and themes

[@ramversingh7867]

Awesome 👌. Thanks for sharing valuable knowledge about the topic.

[@be_happy974]

it's great, hardly find teach code video of PINN

[@akshays6272]

Why there is no collocation loss term in the second example?

[@prefachinho]

Is the Jupyter file of the harmonic oscillator demo available anywhere?

[@CAMLabETHZurich]

All code shown in the lectures is here: github.com/benmoseley/DLSC-2023

[@user-jf8mm5jn9l]

why is the physics loss is 0?

[@thisisharold9066]

I think that is because the undamped spring mass system can be model with *second order homogeneous ordinary differential equation*, y'' + p(x)*y' + q(x)*y = 0. If you model for forced response, E.g. charging response of resistor-inductor-capacitor circuit with 3.3 volts as input, then physics will not be 0. y'' + p(x)*y' + q(x)*y = -3.3.