Laboratoire Sinclair - Actualités

Retrouvez ici quelques actualités de l'IA sur le plateau de Saclay, nos rendez-vous, et quelques vidéos de nos événements publics (séminaires, etc.)

Find here some Saclay AI news, our meetings and some public video footages of our researchers

June 2025

N. Bousquet, V. Le Guen and A. Ribes (EDF/SINCLAIR) co-organize with INRIA and CEA a Research Summer School on Physics-based Machine Learning at EDF Saclay. Follow this link for the program and registration B. Iooss (EDF/SINCLAIR) is an invited speaker at the next GRETSI Research School in Signal and Images Processing. More precisions to come.

April 2025

Several EDF researchers from SINCLAIR will talk and present posters at the SAMO 2025 Conference and the satellite event RT-UQ:
- J. Pelamatti: Application of HSIC-Lasso for high-dimensional feature selection in shapelet-based decomposition
- L. Clouvel: Variance-based importance measures for high-dimensional linear model via Johnson indices: Insights and comparisons
- N. Bousquet: New results on Generalized Hoeffding decomposition of numerical models
- V. Chabridon: Epistemic uncertainty management in risk assessment: connections between robustness and sensitivity analysis tools
- B. Ferrere: Generalized Hoeffding Decomposition for Models with Bernoulli Inputs

March or April 2025

N. Atienza, PhD student (Thales/SINCLAIR), will defend his PhD at Saclay. More precisions to come.

January 2025

Z. Elabib and L. Sasal, both PhD students (TotalEnergies/SINCLAIR), will defend their respective PhD on January, 8th and 9th, Abu Dhabi.

December 2024

Z. Elabib (TotalEnergies/SINCLAIR) gives a talk at ICONIP 2024 conference about "Dynamic reservoir prediction and well control optimization using physics-informed framework" V. Chabridon (EDF/SINCLAIR) gives a talk at MEXICO Days 2024 about "Variance-based importance measures in the context of linear regression: comparative analyses and numerical tests."

L. Sasal (TotalEnergies/SINCLAIR) gives a talk at ICONIC 2024 conference on "TempoKGAT: A Novel Graph Attention Network Approach for Temporal Graph Analysis"

November 2024

L. Sasal (TotalEnergies/SINCLAIR) gave a talk at PRICAI 2024 conference on "TG-PhyNN: An Enhanced Physically-Aware Graph Neural Network Framework for Forecasting Spatio-Temporal Data"

A. Ribes(EDF/SINCLAIR) was a keynote speaker at the NAFEMS France Conference (International Association for the Engineering Modelling, Analysis and Simulation Community). The conference brought together a large number of industrialists to discuss concrete applications in AI and Simulation.

C. Labreuche (Thales/SINCLAIR) is an invited speaker at Rencontres Francophones sur la Logique Floue et ses Applications at Brest. He will speak about the following topic: �Towards the use of a decision model (hierarchical Choquet integrals) in machine learning and image processing�.

L. Sasal (TotalEnergies/SINCLAIR) gave a talk at ECMOR conference about her article "A Graph Neural Network-Based Approach for Complex Reservoirs Simulation Surrogate Modelling"

Z. Elabib (TotalEnergies/SINCLAIR) gave a talk at ECMOR conference about his article "Dynamic reservoir prediction and well control optimization using physics-informed framework"

August 2024

N. Atienza (Thales/SINCLAIR) presented his co-authored work entitled "Cutting the black box - Conceptual interpretation of a latent layer with multi-criteria decision aid" at the International Joint Conference on Artificial Intelligence Melisa Platform (IJCAI 2024)

July 2024

A. Ribes, Nawfal Benchekroun and Theo Delagnes (EDF/SINCLAIR) will present their accepted paper "A Fast Learning-Based Surrogate of Electrical Machines using a Reduced Basis" at the ICML 2024 AI for Science Workshop . Furthemore, a keynote dedicated to the EDF-INRIA Melisa Platform will take place at WANT@ICML 2024.

Armand de Villeroch�, EDF PhD Student @ SINCLAIR-CEREA, won the best Poster Award at the INRIA HARMO 22 Conference in Estonia, for his work on "Neural network surrogates for atmospheric dispersion in built area" (co-authors: Rem-Sophia Mouradi, Vincent Le Guen, Patrick Massin, Alban Farchi, Marc Boquet, Patrick Armand) ; Congrats Armand!

A TotalEnergies/SINCLAIR PhD Student, Amin Dhaou , defended his PhD thesis at Ecole Polytechnique on July 4th. He works on the explainability of time series. Congrats Amin!

June 2024

N. Bousquet (EDF/SINCLAIR) talked about "New results in intepretability of models and algorithms with dependent inputs" obtained by the EDF XAI Team at the ECCOMAS Congress (Lisboa)

May 2024

B. Iooss (EDF/SINCLAIR) co-organized a XAI session of the ENBIS Spring Meeting (Dortmund) dedicated to "Trustworthy Industrial Data Science". Some XAI friends were invited: Nicolas Brunel (CapGemini Invent/Quantmetry/ENSIEE), Salim Amoukou (JPMorgan) and our former EDF PhD student Marouane Il Idrissi (soon at UQAM)

March 2024

Two EDF/SINCLAIR PhD Students, Marouane Il Idrissi and Lucas Meyer, defend their respective PhD theses (4th and 11st of March)! They work on AI interpretability and AI-boosted simulation at large scale, respectively.
M. Il Idrissi: "Development of interpretability methods in machine learning for the certification of AI linked to critical systems"
L. Meyer: "Deep Learning for Numerical Simulation at Scale"
The 8th SINCLAIR Workshop is organized on March, 28th at EDF Lab. It will be opened to members of the three companies.

February 2024

V. Chabridon (EDF/SINCLAIR) organizes an invited session on "Conservative and Robust Strategies for Rare Event Probability Estimation" at the SIAM-UQ 2024 Conference (Trieste). B. Iooss (EDF/SINCLAIR) organizes an invited session on "Robustness Analysis of Uncertainty Quantification to Distribution Uncertainty" at this same conference.
Furthermore V. Chabridon, B. Iooss, M. Il Idrissi, N. Bousquet and J. Pelamatti (EDF/SINCLAIR) present or co-author several talks related to meta-modelling, risk analysis using machine learning and sensitivity / interpretability and robustness analysis:
"Decision Criteria for Risk Analysis in Presence of Two Levels of Input Uncertainty"
"Conservative Assessments of Risk Probabilities Using Surrogates"
"Robustness Analysis of Model Uncertainties Using Fisher Distance: the Case of Truncated Distributions"
"Seeing Perturbed Law-based Robustness Analysis of Epistemic Uncertainties in an Info-gap Framework"
"Gaussian Process Regression: New Hyperparameter Estimation Algorithm for More Reliable Prediction "
"Hidden But Essential Recipes for Successful Gaussian Process Metamodeling to Support Uncertainty Quantification in Numerical Simulation" "HSIC Sensitivity Analysis for Discrete Variables" "Hoeffding Decomposition Revisited"

December 2023

V. Le Guen (EDF/SINCLAIR) was an invited speaker at the MASCOT NUM Workshop on "Physics Informed Learning" (Toulouse)

November 2023

Florence Carton et Louis Verny (TotalEnergies/SINCLAIR) organized the first edition of the Industrial Reinforcement Learning Workshop , an event to gather the RL community and share industrial applications that took place in Saclay.
In addition to SINCLAIR members, this event brought together researchers from several major companies (MBDA, CEA, Valeo), professors from top research pools (Mines Paris and ENSTA Paris), and speakers from foreign institutes such as DLR (the German aerospace center) and Offis - institute for information technology.

L. Meyer and A. Ribes (EDF/SINCLAIR) talked about the article "High Throughput Training of Deep Surrogates from Large Ensemble Runs" accepted at the Supercomputing 2023 Conference (Denver)

V. Chabridon (EDF/SINCLAIR) talked about UQ and ML in Industry, and practices / challenges for low-carbon electricity production, at the "AISSAI Workshop on Artificial Intelligence and the Uncertainty Challenge in Fundamental Physics" (Paris)

M. Riou (Thales/SINCLAIR) presented an article at the Conference on Artificial Intelligence for Defense (Rennes) on the following subject: "Towards simulation of radio-frequency component with physics informed neural networks".

September 2023

B. Iooss (EDF/SINCLAIR) talked about the article "Global sensitivity analysis of model outputs with dependent inputs: New insights around Shapley effects" at the Journées de Géostatistique 2023 (Fontainebleau) He will talk bout the article "Global importance measures for machine learning model interpretability, an overview" at the ENBIS 2023 Conference (Valencia)

A. Dhaou (TotalEnergies/SINCLAIR) presented his work on "Interpretable multi-step ahead time series forecasting" at ECML PKDD 2023

On Tuesday 12 September, the Saclay R&D teams were delighted to welcome a delegation of German researchers from the DFKI (German Research Centre for Artificial Intelligence). The DFKI aims to conduct research on "human-centred AI" in the main innovative areas of AI research and applications, with a focus on socially relevant topics.
Representatives of DFKI projects and start-ups working on Artificial Intelligence were welcomed by three of our R&D staff from TotalEnergies: Sébastien Gourvénec (Catalyst for the Paris-Saclay Ecosystem), Florence Carton (SINCLAIR Researcher and team leader in Reinforcement Learning) and Louis Verny (SINCLAIR interim director and AI researcher). The delegation had the opportunity to visit SINCLAIR and to discover She was able to discover the laboratory's various research areas and identify potential partnership opportunities. In particular, the visit sparked mutual interest in collaborating on two major topics: AI-assisted crisis anticipation and control, and the management of climatic uncertainties through the renovation of existing infrastructures.

July 2023

V. Thouvenot (Thales/SINCLAIR) talked about the article "TSCFKit and CFKit: two Python modules dedicated to counterfactual examples", and B. Iooss (EDF) will present the article "An approximation method for Shapley effects in high dimension" at the Journées de Statistique 2023 (Brussels)

May 2023

The article Training Deep Surrogate Models with Large Scale Online Learning , co-authored by our PhD student L. Meyer and our researcher A. Ribes (EDF), and the article Comparison of meta-learners for estimating multi-valued treatment heterogeneous effects , co-authored by our researcher A. Bertoncello (TotalEnergies), are accepted at ICML 2023!

April 2023

SINCLAIR supported the creation of the group "Epistemological and Ethical Issues in Deep Machine Learning" at the F. Bull Institute

Nicolas Bousquet and Bertrand Iooss (EDF/SINCLAIR) co-organized the 2023 annual meeting of the French CNRS research network MASCOT-NUM , co-chaired by N. Bousquet and E. Vazquez (CentraleSupelec)
A talk was given by M. Il Idrissi (EDF/SINCLAIR) on the following subject: "Coalitional decompositions of quantities of interest: an input-centric point of view"
Several SINCLAIR PhD students presented a poster, including A. Dhaou (TotalEnergies/SINCLAIR) about "Interpretable multi-step ahead time series forecasting".

March 2023

Lucas Meyer, Alejandro Ribes (EDF/SINCLAIR) and their coauthors presented their works on Training Deep Surrogate Models with Large Scale Online Learning at the 15th JLESC Workshop (INRIA Bordeaux, France)

December 2022

Lucas Meyer and Alejandro Ribes (EDF/SINCLAIR) presented their works on Simulation-Based Parallel Training at two NeurIPS Workshops (New Orleans, USA) (see the page "Publications")

November 2022

The first SINCLAIR PhD (TotalEnergies + Ecole Polytechnique) was defended by Naoufal Acharki on November, 22, at Ecole Polytechnique.

Congratulations to Vincent Le Guen (EDF/SINCLAIR), winner of the prestigious Paul Caseau 2022 Prize, awarded by the French Academy of Technologies, in the field of development of electricity uses, energy efficiency and technico-economic analysis of the electrical system, for his doctoral work.

Lucas Meyer and Alejandro Ribes (EDF/SINCLAIR), along with François Mazen, talked about a short paper entitled "In Situ Monitoring and Steering Deep Learning Training from Numerical Simulations in ParaView-Catalyst" at the Conference on In Situ Infrastructures for Enabling Extreme-Scale Analysis and Vizualisation (ISAV'20) at Dallas, TX, USA

Nicolas Bousquet (EDF/SINCLAIR) was a keynote speaker at the MATHMET Conference (Paris), the MEXICO Meeting (Bordeaux) and the MASCOT-NUM Workshop on Statistical Methods for safety and decommissioning (Avignon).

October 2022

Bertrand Iooss, Vincent Chabridon (EDF/SINCLAIR) and Vincent Thouvenot (Thales/SINCLAIR) gave a talk about "Variance-based importance measures for machine learning model interpretability" at the Lambda-Mu (23) Conference at EDF Lab Saclay, France.

May 2022
Open seminar about "Causality and AI explainability" at EDF Lab Saclay

Please register for a remote participation here.

May 2022

Congratulations to Vincent Le Guen (EDF/SINCLAIR) who received an accessit to the thesis prize of the French Association for Artificial Intelligence (AFIA)I and is laureate of the PhD Thesis Prize of the Groupement de Recherche (GdR) ISIS / GRETSI for his work defended at the end of 2021 and entitled "Deep learning for spatio-temporal forecasting - application to photovoltaic energy"

April 2022

Several researchers from EDF / SINCLAIR (B. Iooss, V. Chabridon, N. Bousquet, J. Pelamatti) talked at the SIAM Conference on Uncertainty Quantification (SIAM-UQ) about explainable AI and the virtuous use of machine learning models for stochastic inversion. V. Chabridon will organize a special session on "metamodel-based approaches for robust (stochastic) inversion and optimization".

March 2022

Several researchers from EDF / SINCLAIR (B. Iooss, V. Chabridon, M. Il Idrissi) talked at the 10th International Conference on Sensitivity Analysis of Model Output (SAMO) about explainable AI

October-November 2021

C. Labreuche (Thales R&T, SINCLAIR) was a keynote speaker at the 7th International Conference on Algorithmic Decision Theory
N. Bousquet (EDF R&D, SINCLAIR) was a keynote speaker at the Saclay Research & Innovation Day
B. Iooss (EDF R&D, SINCLAIR) was an invited speaker at the MEXICO Meeting