This training aims at understanding and manipulating the AI tools on both regression and
control tasks encountered in physics. The fundamentals of AI are introduced, first from a
theoretical view point, and second from an application perspective using python notebooks
and open-sources libraries such as ScikitLearn, Keras/Tensforflow and DeepXDE. A large
range of methods of Deep Learning and Deep Reinforcement Learning are considered in this
training, including MLP, CNN, auto-encoders, VAE, PINN, Q-learning etc., each method
being tested during the training on physical datasets (rocket engines, unsteady heat equation,
mechanics). This training also focuses on the specificities of AI applied to physical systems
(e.g. mesh, physics-informed methods etc.).

Objective of the training

The main objective is to have a general theoretical backhground on AI methods, and to apply
them using open-source libraries (ScikitLearn, Keras/TensorFlow, DeepXDE) for both
regression and control tasls encountered in physics.

Learning outcomes

On completion of this course, you will be able to :

– Understand the fondamentals of AI.
– Set a learning problem (regression/control, cost function, hyper-parameters etc.).
– Use some open-sources libraries such as ScikitLearn, Keras/TensorFlow or DeepXDE.
– Evaluate and test trainings involving deep neural networks.

Teaching methods

The training is an alternation of theoretical presentations and practical work. A multiple choice question allows the final evaluation. The training room is equipped with computers, the work can be done in sub-groups of two people.

Referent teacher: Mickaël BAUERHEIM

Target participants

This course is for anyone wishing to process numerical data using modern AI tools.

Prerequisites

• Be an employee of a European company; a certificate from the employer is required
• Have at least 5 years of high education or Master 2 trainee
• Basic knowledge of Python
• Knowledge of general mathematics and physics
• The training can take place in French or English depending on the audience, level B2 of the CEFR is required.

In order to verify that the prerequisites are satisfied, the following questionnaires must be completed. You need to get at least 75% of correct answers in order to be authorized to follow this training session. If you don’t succeed it, your subscription will not be validated. You only have two chances to complete it.

Questionnaire 1 : https://forms.gle/nRmPPtK4RywbiCop7

Questionnaire 2 : https://forms.gle/rLeLyUE457Z7Pnxe8

Registration

I certify that I obtained at least 75% of correct answers, I register

Deadline for registration: 15 days before the starting date of each training

Before signing up, you may wish to report us any particular constraints (schedules, health, unavailability…) at the following e-mail address : training@cerfacs.fr

Fee

This training course, financed as part of the European EuroCC2 project, is free of charge and reserved for employees of European Union member companies. It normally costs 2800 € excluding VAT.

However, your registration is subject to the payment of a deposit of 200 €. This sum will be returned to you at the end of the course if your participation has been effective. If not, it will be retained as compensation for the prejudice caused by leaving people unnecessarily on the waiting list.

Program

From 9h00 to 17h00 (1 h break for lunch)

• Day 1 : Introduction, regression problems, overfitting/underfitting, train/validation/test, regularised problems, introduction to ScikitLearn et application to rocket engines data.
• Day 2 : Introduction to deep learning, MLP, backpropagation, optimization (SGD, RMSprop, ADAM), test MLP on playground. Introduction and application using  Keras/Tensorflow.
• Day 3 : Introduction to CNN, padding and résolution, auto-encoder and VAE. General presentation of generative AI. Application of CNN to the unsteady heat equation.
• Day 4 : Introduction to Physics-Informed Neural Network (PINN), theory and backpropagation, introduction to the open-source code DeepXDE. Application of PINN to the unsteady heat equation. Application of DeepXDE to inverse problems.
• Day 5 : Introduction to reinforcement learning (RL), Bellman equation, Q-learning, presentation of RL applications in physics. Use of deep RL on simple mechanical problems. Final exam

Final examination

A final exam will be conducted during the training.