Différences entre les versions de « MINI-SCHOOL 2025 »
| (4 versions intermédiaires par 2 utilisateurs non affichées) | |||
| Ligne 50 : | Ligne 50 : | ||
* Dimensionality reduction: linear methods (principal component analysis) and nonlinear ones (autoencoders) | * Dimensionality reduction: linear methods (principal component analysis) and nonlinear ones (autoencoders) | ||
* Application: finding collective variables in molecular dynamics | * Application: finding collective variables in molecular dynamics | ||
| + | |||
| + | |||
| + | <font color="red"> | ||
| + | '''Download [[Media:Notes_de_cours_Gabriel_Stoltz.pdf|lecture notes]]''' | ||
| + | </font> | ||
| Ligne 70 : | Ligne 75 : | ||
'''References:''' | '''References:''' | ||
| − | The main reference for the first course is the introductory book by Kevin Murphy, which focuses on algorithms and numerical methods. It covers a wide range of methods and techniques. The book can be found here: | + | The main reference for the first course is the introductory book by Kevin Murphy, which focuses on algorithms and numerical methods. It covers a wide range of methods and techniques. The book can be found here: https://probml.github.io/pml-book/book1.html |
| + | |||
| + | Another nice reference for a public familiar with statistical physics is: P. Mehta, M. Bukov, C.-H. Wang, A.G.R. Day, C. Richardson, C.K. Fisher, D.J. Schwab, A high-bias, low-variance introduction to Machine Learning for physicists, Physics Reports 810, 1-124 (2019) https://arxiv.org/abs/1803.08823 | ||
| − | + | Other nice references (occasionally mentioned in the course) include: | |
| + | - S. Shalev-Shwartz and S. Ben-David, Understanding Machine Learning https://www.cs.huji.ac.il/~shais/UnderstandingMachineLearning/understanding-machine-learning-theory-algorithms.pdf | ||
| − | + | - D. Barber, Bayesian Reasoning and Machine Learning http://web4.cs.ucl.ac.uk/staff/D.Barber/textbook/090310.pdf | |
| − | + | - C. Bishop, Pattern Recognition and Machine Learning | |
| + | - E. Alpaydin, Introduction to Machine Learning | ||
| + | |||
| + | - M. Mohri, A. Rostamizadeh and A. Talwalkar, Foundations of Machine Learning https://cs.nyu.edu/~mohri/mlbook/ | ||
| + | |||
| + | - F. Bach, Learning Theory from First Principles, MIT Pres https://www.di.ens.fr/~fbach/ltfp_book.pdf | ||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
== Schedule == | == Schedule == | ||
Dernière version du 6 mai 2025 à 07:35
(organized by GDR NBODY with support from CNRS Chimie, LCT, LJLL, and ERC EMC2)
In the same spirit than the 1st edition, 2nd edition, 3rd edition, 4th edition, 5th edition, and 6th edition
- Dates: 19-21 May 2025
- Location: Sorbonne Université, Pierre et Marie Curie (or Jussieu) campus, 4 place Jussieu, 75005, Paris. Laboratoire Jacques-Louis Lions, tower/corridor 15-16, 3rd floor, room 309.
- Lecturers: Geneviève Dusson and Gabriel Stoltz.
- Organizers: Eric Cancès, Mi-Song Dupuy, Emmanuel Fromager, Pierre-François Loos, Julien Toulouse. contact email
Presentation[modifier]
The lectures delivered are intended to be of interest to any person working in the field of theoretical chemistry and physics and willing to discover or deepen some mathematical aspects of the field. Master and PhD students, post-docs and any academic are welcome!
Program[modifier]
Two courses will be proposed:
1 - (Un)supervised learning with applications to molecular dynamics (9 hours) by Gabriel Stoltz (École des Ponts ParisTech)
- What is machine learning? A survey of supervised and unsupervised learning
- Regression in the context of supervised learning
- K-nearest neigbors: a first method to introduce various concepts (excess risk, empirical risk, cross validation, ...)
- linear least-square predictors and the need for regularization to prevent overfitting (ridge/LASSO)
- kernel methods
- trees and ensemble methods
- neural network approaches
- a word on the training of machine learning models (SGD/Adam, minibatching, ...)
- Unsupervised learning
- Clustering
- Dimensionality reduction: linear methods (principal component analysis) and nonlinear ones (autoencoders)
- Application: finding collective variables in molecular dynamics
Download lecture notes
2 - From machine-learned wavefunctions to interatomic potentials (9 hours) by Geneviève Dusson (CNRS and Université Franche-Comté)
- From wavefunctions to interatomic potentials: What is to be learned
- Quick summary of the theory, Schrödinger equation, Wavefunction, Electronic density, Density matrix, Potential energy surface
- Data, Functionals to approximate, Objective/cost functions
- Atomic descriptors: Dealing with the symmetries, Invariance and equivariance, Review of existing descriptors
- Linear methods (MTP, PIPs, ACE, …) & Kernel methods (GAP)
- Neural networks for interatomic potentials & MACE (tensor network)
- ML wavefunctions: PINNs + PauliNet, FermiNet
References:
The main reference for the first course is the introductory book by Kevin Murphy, which focuses on algorithms and numerical methods. It covers a wide range of methods and techniques. The book can be found here: https://probml.github.io/pml-book/book1.html
Another nice reference for a public familiar with statistical physics is: P. Mehta, M. Bukov, C.-H. Wang, A.G.R. Day, C. Richardson, C.K. Fisher, D.J. Schwab, A high-bias, low-variance introduction to Machine Learning for physicists, Physics Reports 810, 1-124 (2019) https://arxiv.org/abs/1803.08823
Other nice references (occasionally mentioned in the course) include:
- S. Shalev-Shwartz and S. Ben-David, Understanding Machine Learning https://www.cs.huji.ac.il/~shais/UnderstandingMachineLearning/understanding-machine-learning-theory-algorithms.pdf
- D. Barber, Bayesian Reasoning and Machine Learning http://web4.cs.ucl.ac.uk/staff/D.Barber/textbook/090310.pdf
- C. Bishop, Pattern Recognition and Machine Learning
- E. Alpaydin, Introduction to Machine Learning
- M. Mohri, A. Rostamizadeh and A. Talwalkar, Foundations of Machine Learning https://cs.nyu.edu/~mohri/mlbook/
- F. Bach, Learning Theory from First Principles, MIT Pres https://www.di.ens.fr/~fbach/ltfp_book.pdf
Schedule[modifier]
| Monday 19 May
Tower 15-16, room 309 |
Tuesday 20 May
Tower 15-16, room 309 |
Wednesday 21 May
Tower 15-16, room 309 |
|---|---|---|
| 8:45-9:00 Arrivals/Welcome | ||
| 9:00-10:30 Gabriel Stoltz 1 | 9:00-10:30 Gabriel Stoltz 3 | 9:00-10:30 Geneviève Dusson 5 |
| 10:30-10:45 Coffee break | 10:30-10:45 Coffee break | 10:30-10:45 Coffee break |
| 10:45-12:15 Gabriel Stoltz 2 | 10:45-12:15 Gabriel Stoltz 4 | 10:45-12:15 Geneviève Dusson 6 |
| 12:15-14:00 Lunch at L'Ardoise (Tower 25) | 12:15-14:00 Lunch at L'Ardoise (Tower 25) | 12:15-14:00 Lunch at L'Ardoise (Tower 25) |
| 14:00-15:30 Geneviève Dusson 1 | 14:00-15:30 Geneviève Dusson 3 | 14:00-15:30 Gabriel Stoltz 5 |
| 15:30-15:45 Coffee break | 15:30-15:45 Coffee break | 15:30-15:45 Coffee break |
| 15:45-17:15 Geneviève Dusson 2 | 15:45-17:15 Geneviève Dusson 4 | 15:45-17:15 Gabriel Stoltz 6 |
Practical details[modifier]
Registration[modifier]
The registration is closed.
To facilitate the participation of young researchers, no registration fees will be asked. However, accommodation will not be provided but must be arranged independently by the participants.
Location of the mini-school[modifier]
The mini-school will take place on the Pierre et Marie Curie (or Jussieu) campus of Sorbonne Université, 4 place Jussieu, 75005 Paris.
Subway (Metro) lines 7 or 10: get off at "Jussieu" station.
From the main entrance of the campus, go to Tower 16. Take the stairs or the elevator to the third level. Find corridor 15-16, and then the room 309. This is the seminar room of Laboratoire Jacques-Louis Lions. All the lectures will take place in this seminar room.
Lunches[modifier]
Lunches at Restaurant L'Ardoise (Jussieu campus, entrance near Tower 25) will be offered to the first 50 registered participants to the mini-school (see the list of registered participants on the page List of registered participants).
Supported by 
