IA_CNRS_ICA

Noteboook and Data for ICA's course on IA

Part 0. A very good introduction by Prof. Sid Kumar

Machine Learning Applications in Material Science - Part 1

Machine Learning Applications in Material Science - Part 2

Notebooks will come

Verified on colab

Part 1. Gaussian processes Regression

Introduction to Gaussian processes, mathematical foundations in 30’ Exercise: How can I program my GPR 30’ Beware of noisy data when using sigma_n as noise !

Part 2. Gaussian processes versus neural networks

Inspired by Prof. Miguel Bessa Imechanica Large-scale illustration, hyper parameter tuning, BO in 30’ Exercise: How do I use toolboxes on more complex examples 45’

Part 3. Physics Informed Neural Networks {PINN}

What is a PINN knowing a NN? in 15' Exercise: How to Train a PINN to simulate a dynamic system (damped harmonic oscillator) in 30’s

Solution in video

Bonus: Train a PINN to reverse the underlying parameters

in french

Part 1. Régression par processus gaussiens

Introduction aux processus gaussiens, fondements mathématiques en 30’ Exercice: Je programme mon GPR 30’ Attention bruiter les données quand on utilise sigma_n comme bruit

Part 2. processus gaussiens versus réseaux de neurones

Illustration en grande dimension, hyper parameter tuning, BO en 30’ Exercice: J’utilise des toolboxes sur des exemples plus complexes 45’

Part 3. Physics Informed Neural Networks {PINN}

Qu’est ce qu’un PINN connaissant un NN? en 15’ Exercice: Entrainer un PINN pour simuler le système (oscillateur harmonique amorti) en 30’

Bonus: Entrainer un PINN pour inverser les paramètres sous-jacents

Aknowledgments in the notebooks

Online references

https://neurips.cc/virtual/2021/tutorial/21890

http://www.infinitecuriosity.org/vizgp/

https://distill.pub/2019/visual-exploration-gaussian-processes/