Machine learning for physics and engineering

25

English

Mines Paris;

This week of lectures aims to share data and algorithms for machine learning applications in physics and model-based engineering. ;
Monday : ;
Introduction of the lectures, main motivations for the developpement of machine learning and artificial intelligence in physics and engineering, especially in model-based engineering. Two parallel sessions are proposed depending on the background of students. The session « Numerical modeling » presents the data in physics and in model-based engineering. The session « Machine learning » is an introduction to supervised and unsupervised machine learning, regressor, classifiers and stochastic gradient algorithm.
Tuesday :
Lecture and exercices on deep learning for computer vision, via convolutional neural networks (CNN).
Wednesday :
Lectures and exercices on complexity reduction of physical models in engineering by using computer vision. On the morning, we will consider data structured on graphs and related Graph Neural Networks. On the afternoon, deep classifiers are presented for the recommandation of digital twins having a reduced setting. ;
Thursday :
Recurrent neural networks, transformers, for natural langage processing and many other applications related to time series. ;
Friday :
Reinforcement learning is applied to computational fluid dynamics for process optimisation in industry, and an exam for attendees that need a mark to complete master courses.

Monday 9:00-10:00, D. Ryckelynck;
Introduction;
Monday 10:15-12:00, D. Ryckelynck, group A;
Numerical modeling in model based-engineering;
Monday 10:15-12:00, Arturo Amor, group B;
Machine learning;
Monday 14:00-15:30, Elie Hachem, group A;
Computational Fluid mechanics;
Monday 15:45-17:00, David Ryckelynck, group A;
Digital Twining;
Monday 14:00-17:00, Arturo Amor, group B;
Machine Learning;
Tuesday 9:00-10:30, Matteo Bastico;
Deep learning for computer vision;
Tuesday 10:45-12:00, Matteo Bastico;
Data challenge on image segmentation;
Tuesday 14:00-17:00, Matteo Bastico;
Data challenge on image segmentation;
Wednesday 9:00-10:30, Pierre Kerfriden;
Graph Neural Networks;
Wednesday 10:45-12:00, Pierre Kerfriden;
Numerical exercises on GNN;
Wednesday 14:00-17:00, David Ryckelynck;
Deep classifiers of digital twins;
Thursday 9:00-12:00, Matthieu Labeau;
Natural Language Processing;
Thursday 14:00-17:00, Matthieu Labeau;
Exercises on Natural Language Processing;
Friday 9:00-12:00 David Ryckelynck;
Auto-encoders for manifold learning and model reduction;
Friday 14:00-15:30, Jonathan Viquerat;
Deep Reinforcement Learning;
Friday 15:30-17:00;
Exam;

Quiz.

Python programming.

RYCKELYNCK David

D. Ryckelynck, E. Hachem, M. Bastico, M. Labeau, P. Kerfriden, J. Viquerat.