About
The summer school is targeted toward PhD students working with data science and/or AI broadly, and for whom generative modeling potentially plays a part in their projects. In particular, the program is designed to accommodate both students doing methodological data science research (e.g., machine learning, statistics, and AI) and students doing applied data science research (e.g., bioinformatics, computational physics, computational chemistry, and computational social science). Furthermore, the course will also be open to postdocs and more senior researchers from the industry as continual training in AI.
This year, GeMSS will be colocated with Statlearn. Statlearn is a scientific workshop held every year, which focuses on current and upcoming trends in statistical learning. Statlearn is a scientific event of the French Society of Statistics (SFdS) that has been organised since 2010. Conferences and spring school are organised alternatively every other year.
Lecturers
The first three days will be taught by the main lectures, while there will be invited lectures giving tutorials the last two days.
Invited lecturers
 
Main lectures
 
Program
Monday (Mar 31) | Tuesday (Apr 1) | Wednesday (Apr 2) | Thursday (Apr 3) | Friday (Apr 4) | |||||
---|---|---|---|---|---|---|---|---|---|
Day 1 Lectures & Hands-on |
Day 2 Lectures & Hands-on |
Day 3 Lectures & Poster session |
Day 4 Invited Lectures & Dinner |
Day 5 Invited Lectures |
|||||
9:00-9:15 | Opening remarks Registration from 8:15 Charles Bouveyron, Pierre-Alexandre Mattei |
||||||||
9:15-10:30 | Introduction to generative modeling Pierre-Alexandre Mattei Ch. 1 |
9:30-10:30 | Autoregressive models (Introduction) Jakub Tomczak Ch. 3 |
9:30-10:30 | From LLMs to GenAISys Jakub Tomczak Ch. 11 |
09:00-10:30 | Inverting Scientific Images With Score-Based Generative Models Gilles Louppe |
09:00-10:30 | Latent space (generative) models in recent scientific breakthroughs Johannes Brandstetter |
11:00-12:00 | From Mixture Models to Probabilistic circuits Jakub Tomczak Ch. 2 |
11:00-12:00 | Autoregressive models (Transformers) Jakub Tomczak Ch. 3 |
11:00-12:00 | Energy-based models Jes Frellsen Ch. 7 |
11:00-12:30 | Sequential Latent Variable Models Yingzhen Li |
11:00-12:30 | Implicit Diffusion: Efficient Optimization through Stochastic Sampling Anna Korba |
12:00-13:30 | Lunch |
12:00-13:30 | Lunch |
12:00-13:30 | Lunch |
12:30-14:00 | Lunch |
12:30-14:00 | Lunch |
13:30-15:00 | Probabilistic PCA Pierre-Alexandre Mattei Ch. 5.2 |
13:30-15:00 | Flow-based models Jes Frellsen Ch. 4 |
13:30-15:00 | Diffusion-based generative models Jakub Tomczak Ch. 9 |
14:00-14:45 | Biases in LLMs Frédéric Precioso |
14:00-15:30 | Subtractive Mixture Models: Representation, Learning and Inference Antonio Vergari |
15:30-17:00 | Deep latent variable models Jes Frellsen Ch. 5 |
15:30-16:30 | Flow-based models in VAEs, Anomaly detection using generative models Jes Frellsen, Pierre-Alexandre Mattei Ch. 5.4.2 |
15:30-16:30 | Generative Adversarial Networks Pierre-Alexandre Mattei Ch. 8 |
15:00-16:30 | Generative Methods for Missing Data: From PCA to Deep Learning Aude Sportisse |
16:00-16.30 | Closing Pierre-Alexandre Mattei |
17:15-18:30 | Hands-on session VAE |
17:00-18:30 | Hands-on session ARM/Flows |
17:00-19:00 | Poster session |
16:45-17:30 | Bridging generative models and Neural Networks for domain-agnostic segmentation of brain MRI Benjamin Billot |
||
19:00 | Gala dinner Bistrot Margaux + Da Vinci trattoria, Antibes |
Recommended reading
We recommend Deep Generative Modeling by Jakub M. Tomczak (Second Edition, 2024, Springer) as background reading for the summer school. The book provides a comprehensive introduction to the key concepts and methods covered in the lectures. For each lecture, the corresponding chapter from the book is indicated in the program above (e.g., Ch. 1).
Venue
The Summer School will be hosted at the Inria Centre at Université Côte d’Azur (Sophia Antipolis, France). Sophia Antipolis is a large tech park located in a forest in the French Riviera, near the cities of Nice, Antibes, and Cannes.
The most convenient option will be to choose an accomodation in Antibes, since public transportation between Antibes and Sophia Antipolis is quite efficient. It is also possible to stay in Sophia Antipolis or in one of the nearby villages like Biot. There is also a bus between Nice and Sophia Antipolis. For more details on how to reach the Inria campus, see here.
Submission & Registration
Application
Please prepare a single PDF file that contains the following information in the given order:
- one A4 page describing your research, preferably in a poster format;
- a one-page CV;
- students only: a one-page letter of confirmation that you are a PhD student from your supervisor;
Please upload this single file to the Statlearn website (please create a submission, do not register yet).
An example of an application could be found here: [PDF].
Application deadline: January 27, 2025 (23:59 Central European Summer Time). The deadline has been extended to February 2, 2025 (23:59 Central European Summer Time).
Selection Process
We aim for selecting a group of about 150 participants (the max. capacity of lecture rooms). The selection process will be based on the submitted material and will be executed by the organisers. We will look into your background (e.g., expertise, relevance) and your experience. We expect from the participants the following: a good familiarity with Python and PyTorch, and a basic knowledge of calculus, linear algebra, probability theory and statistics.
Registration Fee
- PhD students: 400EUR
- Academic participants: 500EUR
- Industrial participants: 800EUR
The registration fee includes participating in all lectures, coffee breaks, lunches and the banquet. All other costs not specified in the previous sentence (e.g., accomodation, travels, communication, other meals) are NOT provided by the organisers.
Code of conduct
An environment that recognises and respects the inherent worth of every person is crucial for a successful summer school. We, therefore, expect everyone engaged in the summer school to contribute to a safe and inclusive environment and respect the guidelines stipulated below.
This Code of Conduct applies to all participants, including organisers, volunteers, speakers/teachers.
Offensive behaviour
The course must provide an experience for all participants that is free from harassment, bullying, discrimination and inappropriate behaviour, which includes, but is not limited to:
- Intimidation, personal attacks, stalking, following, harassment, inappropriate physical contact or attention, or derogatory mention
- Offensive comments, materials or activities related to age, race, colour, nationality, religion, gender/gender identity, sexual orientation, medical condition, disability, or pregnancy
- Disorderly, boisterous, or disruptive conduct, including fighting, coercion, theft, damage to property, or any mistreatment or unprofessional behaviour towards participants
Procedure
Any participant in summer school may contact us about experiences or observations of inappropriate behaviour, regardless of whether they wish to place a formal complaint or merely express a concern or seek advice. Please contact any of the Named Persons, Chloé Friguet or Jes Frellsen, directly at the event or by email.
- The Named Person will support the person reporting an offensive behaviour in deciding if and how to proceed.
- The Named Person has no formal authority but will support the person reporting offensive behaviour by informing the relevant authority if agreed upon.
- The Named Person is not obliged to inform any other authority (unless required by law) and is thus able to handle any complaint or concern in full confidentiality.
Action
In case of a breach of this Code of Conduct, the organiser will respond towards the offender as deemed appropriate according to law and regulations.
Sanctions may be, but are not limited to, a warning, expulsion from the course with no refund, reporting the incident to the offender’s organisation/company, or, in the last instance, reporting the incident to law enforcement authorities.
With permission, this Code of Conduct has been adapted from the DDSA Code of Conduct.
Organisation
GeMSS committe:
- Jes Frellsen (Technical University of Denmark)
- Pierre-Alexandre Mattei (Inria, Université Côte d'Azur)
- Jakub M. Tomczak (Eindhoven University of Technology)
Statlearn committe:
- Charles Bouveyron (Université Cote d’Azur, Inria)
- Irene Balleli (Inria, Université Côte d'Azur)
- Marco Corneli (Université Cote d’Azur)
- Mathieu Emily (Agrocampus Ouest)
- Chloé Friguet (Univ. Bretagne-Sud)
- Julien Jacques (Univ. Lyon)
- Pierre Latouche (Univ. Clermont-Auvergne)
- Pierre-Alexandre Mattei (Inria, Université Côte d'Azur)
Contact
For matters regarding the summer school, please contact .