ETICS Annual Research Schools
École Thématique sur les Incertitudes en Calcul Scientifique
Research School on Uncertainty in Scientific Computing
Organized by CEA/DAM, EDF R&D and ENS Paris Saclay (Centre Borelli andLRC MESO), under the collaborative actions of the GIS LARTISSTE and with the scientific labelling of the GdR MASCOT-NUM
Scientific committee: Claire Cannamela (CEA/DAM), Sébastien Da Veiga (SafranTech), Bertrand Iooss (EDF R&D), Merlin Keller (EDF R&D), Amandine Marrel (CEA/DES), Guillaume Perrin (Université Gustave Eiffel), Gael Poette (CEA/DAM)
The goal of this school is to develop the skills of researchers and engineers in the domain of uncertainty management of computer codes and machine learning techniques in support to engineering studies. Some of the lectures are followed by practical computer works. Collaborative works, round tables and poster sessions allow to promote exchanges between participants. The prerequisites to possess are the mathematical bases of the uncertainty quantification science.
ETICS2023
Dates: 2023, October, 8th-13th
Location: VVF Lège Cap Ferret, France
Warning: The language of this ETICS edition will be in French - Les cours de cette édition d'ETICS seront dispensés en français
Organizers : Bertrand Iooss (EDF R&D) and Claire Cannamela (CEA/DAM)
Lecturers:
- Prof. Pierre Barbillon (AgroParisTech): Calibration of deterministic and stochastic computer models
- Prof. Sébastien Le Digabel (Polyetchnique Montréal): Black-box optimization
- Prof. Mathilde Mougeot (ENSIIE and ENS Paris Saclay): Leveraging knowledge to design machine learning despite the lack of data using Transfer Learning and PINNS models
Talks from PhD students: TBA
Registration
Registration is closed because the maximal number of participants has been reached. - Contact Bertrand Iooss to be on the waiting list
Registration fees (935€, taxes included, 850€ without taxe) include accommodation, meals and transport by bus from and to Bordeaux train station.
access to the list of participants
Abstracts
- Prof. Pierre Barbillon (AgroParisTech): Calibration of deterministic and stochastic computer models - We propose an overview of calibration methods for deterministic computer models. The choice of the design of experiments and the validation of computer models will be dealt with. The extension to stochastic simulators will be discussed.
- Prof. Sébastien Le Digabel (Polyetchnique Montréal): Black-box optimization - Derivative-Free Optimization typically occurs in engineering applications, when the function to be optimized, as well as the constraints of the problem, are given by the outputs of a computer simulation seen as a blackbox, possibly expensive to execute and contaminated by noise. In such a context, the derivatives are neither accessible nor calculable, which prevents the application of classical gradient-based methods. This course adopts a practical approach to solve such problems, by going around the different families of methods, as well as the different software available. Specific subjects will also be covered, such as handling of constraints, use of surrogates, discrete variables, multiobjective optimization, etc. The practical course includes manipulation of a solver as well as the modeling of a problem in the blackbox optimization setting.
- Prof. Mathilde Mougeot (ENSIIE and ENS Paris Saclay): Leveraging knowledge to design machine learning despite the lack of data using Transfer Learning and PINNS models - In recent years, considerable progresses have been made in the implementation of decision support procedures based on machine learning methods through the exploitation of very large databases and the use of learning algorithms. In many research or production environments, the available databases are rarely so voluminous and the question arises as to whether in this context it is reasonable to use machine learning methods. The ETICS lessons will introduce transfer learning and hybrid models that use knowledge from related application or from physics to implement efficient models with an economy of data. Several achievements will be presented that successfully use these learning approaches to design machine learning for industrial small data regimes and to develop powerful decision support tools even in cases where the initial data volume is limited. Practical sessions will help to handle such models in python environments. References: -de Mathelin, A., Deheeger, F., Mougeot, M., Vayatis, N. From Theoretical to Practical Transfer Learning: the ADAPT library, Federated and Transfer Learning Springer book. 2023. -Weiss K., Khoshgoftaar, T Wang, D. A survey of transfer learning, Journal of Big data, 2016. -Nguyen, Khoa.T.N, Dairay, T., Meunier, R., Mougeot, M. (2022) Physics-informed neural networks for non-Newtonian fluid thermo-mechanical problems: an application to rubber calendering process, Engineering Applications of Artificial Intelligence, vol. 141. -Raissi, M., Perdikaris, P. Karniadakis, G.E. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics, 2019.
- PhD students' abstracts: TBA
ETICS2022
Dates: 2022, October, 2nd-7th
Location: Belhambra Hotel, Golfe de Lozari (L'Ile Rousse), France
Organizers : Bertrand Iooss (EDF R&D) and Claire Cannamela (CEA/DAM)
Lecturers:
- Anthony Nouy (Ecole Centrale de Nantes): High-dimensional approximation - part 1 - part 2 - part 3
- Gabriel Peyré (CNRS and Ecole Normale Supérieure): Computational Optimal Transport - link - Course notes - Slides 1 - Slides 2 - Python codes (section « Optimal Transport », 5 Jupiter notebooks)
- Danica J. Sutherland (University of British Columbia and Amii): Modern Kernel Methods in Machine Learning - link to the interactive material
- Delphine Sinoquet (IFPEN): - Two applicatins of stepwise uncertainty reduction strategy for excursion set estimation - etics22_sinoquet.pdf
- Clément Gauchy, Estimation of seismic fragility curves by sequential design of experiments (third year PhD) - etics22_gauchy.pdf
- Clément Hardy, Off-the grid learning of sparse mixtures from a continuous disctionary (third year PhD) - etics22_hardy.pdf
- Inês Cardoso, Model Order Reduction and Bayesian Optimization for MDO problems - etics22_cardoso.pdf
- Guillaume Chennetier, Adaptive importance sampling for reliability assessment of a Piecewise Deterministic Markov Process - etics22_chennetier.pdf
- Julien Demange-Chryst, Shapley effect estimation in reliability-oriented sensitivity analysis with dependent inputs by importance sampling - etics22_demange.pdf
- Elias Fekhari, Sequential reliability analysis for offshore wind turbine fatigue Assessment - etics22_fekhari.pdf
- Noé Fellmann, Sensitivity Analysis on excursion sets - etics22_gauchy.pdf
- Marouane Il Idrissi, Robustness assessment using quantile-constrained Wasserstein projections - etics22_ilidrissi.pdf
- Matthéo Saldanha, Using Generative Adversarial Networks to constrain inverse problems resolution - etics22_saldanha.pdf
- Babacar Sow, Gaussian processes indexed by clouds of points: a study - etics22_sow.pdf
- Charles Surget, Sensitivity to statistical estimation uncertainties and probabilistic model identification - etics22_surget.pdf
- Olivier Truffinet, Exploration and compression of homogeneized cross-sections by the EIM method - etics22_truffinet.pdf
Registration
Pre-registration for ETICS2022
The registration is closed because the maximal number of participants have been reached.
Abstracts
- Anthony Nouy (Ecole Centrale de Nantes): High-dimensional approximation, tensor networks and beyond - The approximation of high-dimensional functions is a typical task in computational science. Examples of such problems can be found in physics, machine learning and uncertainty quantification. These tasks require the introduction of suitable approximation tools (or model classes) that are able to exploit some specific structures of functions. In this course, we will first introduce some concepts from approximation theory and information based complexity to quantify the best we can expect when we try to approximate functions from some function class using linear or nonlinear approximation tools, and using different types of information. Then we will review some classical high-dimensional approximation tools, including sparse approximation, dimension reduction methods, tensors and neural networks. We will finally focus on tree tensor networks, provide theoretical results in approximation and learning, and present some algorithms for solving approximation or learning tasks using these model classes.
- Gabriel Peyré (CNRS and Ecole Normale Supérieure): Computational Optimal Transport - Optimal transport (OT) is a fundamental mathematical theory at the interface between optimization, partial differential equations and probability. It has recently emerged as an important tool to tackle a surprisingly large range of problems in data sciences, such as shape registration in medical imaging, structured prediction problems in supervised learning and training deep generatsie networks. This course will interleave the description of the mathematical theory with the recent developments of scalable numerical solvers. This will highlight the importance of recent advances in regularized approaches for OT which allow one to tackle high dimensional learning problems. Material for the course (including a small book, slides and computational resources) can be found online at this link.
- Danica J. Sutherland (University of British Columbia and Amii): Modern Kernel Methods in Machine Learning - Kernel methods were previously the dominant paradigm of machine learning, but have somewhat fallen out of favor with the advent of deep learning. This course will demonstrate how insights from kernel methods can still be beneficial in modern deep learning settings, assuming no prior knowledge of kernels. We will begin with basic formulations of kernel spaces and learning with kernels, including their core theoretical justifications, and with a particular focus on kernel mean embeddings of distributions for various applications. We will then discuss the end-to-end learning of kernels within a deep learning framework, including when we expect this to be useful, with practical examples worked out in interactive sessions.
- PhD students' abstracts: etics22_phdabstracts.pdf
ETICS2021
Dates: 2021, September, 12th - 17th
Location: Keravel resort, Erdeven, France - link
Organizers : Bertrand Iooss (EDF R&D) and Claire Cannamela (CEA/DAM)
Lecturers:
- Prof. Bertrand Michel (Ecole Centrale de Nantes): Topological data analysis - link
- Prof. Chris Oates (Newcastle University): Minimum Discrepancy Methods in Uncertainty Quantification - lecture notes
- Prof. Clementine Prieur (Université Grenoble Alpes): Recent advances in global sensitivity analysis - parti_prieur_etics_2021.pdf - partii_prieur_etics_2021.pdf
- Sébastien Da Veiga (SfranTech) - Kernel-based ANOVA decomposition and Shapley effects - etics2021_daveiga.pdf
- Baptiste Kerleguer (CEA/DAM, 3rd year PhD student): Multi-fidelity surrogate modeling combining Bayesian neural network and Gaussian process regression - etics2021_kerleguer.pdf
- Clément Gauchy (CEA/DES, 2nd year PhD student): Propagation of epistemic uncertainties and global sensitivity analysis in seismic risk assessment - etics2021_gauchy.pdf
- Guillaume Chennetier (EDF R&D, 1st year): Rare event simulation for piecewise deterministic Markov processes - etics2021_chennetier.pdf
- Clément Duhamel (INRIA, 1st year PhD student): A SUR version of the Bichon criterion for excursion set estimation - etics2021_duhamel.pdf
- Elias Fekhari (EDF R&D, 1st year PhD student): Treatment of uncertainties in multi-physics model for wind turbine asset management - etics2021_fekhari.pdf
- Bruno Vuillod (CEA/DAM, 1st year PhD student): Modeling multi-scales thermo-mechanical model - etics2021_vuillod.pdf
- Marouane Il Idrissi (EDF R&D, 1st year PhD student): Cooperative game theory and global sensitivity analysis - etics2021_ilidrissi.pdf
- Isabelle Abraham (CEA/DAM): Maching learning et image : Utilisation du "scattering wavelet transform" sur des radiographies de confinement inertiel - etics2021_abraham.pdf
- Clément Benard (SafranTech, 3rd year PhD student): A sensitivity analysis perspective of random forests - etics2021_benard.pdf
- Cécile Haberstich (CEA/DAM): Optimal designs for discrete least squares approximation - etics2021_haberstich.pdf
- Alejandro Ribes (EDF R&D): Melissa: a modular external library for in-situ sensitivity analysis - etics2021_ribes.pdf
Registration
Registration is open: fill the form below. Registered people will receive the payment detail by email. Registration fees (900€) contain the accommodation, meals, gala dinner, social event cost and transport by bus from and to the railway station (to be precised) to and from Keravel resort (sunday afternoon and friday afternoon). People from CEA/DAM has to precise CEA/DAM for their affiliation.
Registration is closed
ETICS2020, October, 4-9, Ile d&
Organizers : Bertrand Iooss (EDF R&D) and Guillaume Perrin (CEA/DAM)
Location: La Vieille Perrotine, Saint-Pierre-d'Oléron, France - link
Registration is closed
Registration: (900€ including accomodation and meals): French link - English link
Registration is closed
Scientific organizers : Bertrand Iooss (EDF R&D) and Guillaume Perrin (CEA/DAM)
Lecturers:
- Prof. Josselin Garnier (Ecole Polytechnique, France) - Rare event simulation - Slides: etics2020_garnier.pdf - etics2020_garnier2.pdf
- Prof. Anne-Laure Fougères (Université Claude Bernard Lyon 1, France) - Extreme value theory and applications
- Prof. Robert B. Gramacy (Virginia Tech, USA) - Surrogates: Gaussian process modeling, design and optimization for the applied sciences - Part 1: Gaussian processes - lect2_gp.doc (change the file extension .doc to .html) - Part 2: Optimization - lect3_optim.doc (change the file extension .doc to .html) - R files for practical works: r-gramacy.zip
- Baptiste Kerleguer (CEA) : Multi-fidelity modeling for time-series output - etics2020_kerleguer.pdf
- Vincent Chabridon (EDF R&D) : Epistemic uncertainty management in uncertainty quantification etics2020_chabridon.pdf
- Bertrand Iooss (EDF R&D) : Perturbed-law based sensitivity indices : motivations, methodology and industrial application etics2020_iooss_pli.pdf
- Guillaume Perrin (CEA) : Isoprobabilist transform for robust uncertainty analyses etics2020_perrin.pdf - notebook-etics2020-tprobustanalysis.doc (change the file extension .doc to .ipynb)
- Clément Gauchy (CEA) and J. Stenger (EDF R&D) : Optimal Fisher-based perturbed-law indices - etics2020_gauchy.pdf
- Jérôme Breil, Lucas Tallois (CEA) : Optimisation d'un diaphragme hydraulique
- Sanaa Zannane (EDF R&D) : Optimisation coûteuse boîte noire à variables mixtes. Revue bibliographique - etics2020_zannane.pdf
- Rodolphe Le Riche (CNRS) : Dimension reduction ofr Bayesian optimization - slides
- Sébastien Da Veiga (SafranTech) and Amandine Marrel (CEA) : Gaussian process regression with linear inequality constraints – An adaptive strategy - etics2020_daveiga.pdf
- Gaêl Poette (CEA) : A new MC scheme for the resolution of intrusive-gPC based reduced models for the uncertain linearBoltzmann equation - etics2020_poette.pdf
ETICS2019, September, 22-27, Fréjus, France
Organizers : Bertrand Iooss (EDF R&D) and Guillaume Perrin (CEA/DAM)
Location: La Villa Clytia, Fréjus, France - link
Registration (900€ including accomodation and meals): link
Scientific organizers : Bertrand Iooss (EDF R&D) and Guillaume Perrin (CEA/DAM)
Lecturers:
- Prof. Aurélien Bellet (INRIA Lille - Nord Europe, France) - Similarity and distance metric learning - etics2019abt.pdf - metric-learn-etics.zip
- Prof. Bernard Bercu (Université de Bordeaux, France) - Asymptotic behavior of stochastic algorithms with statistical applications - L1-Martingales.pdf - L2-Stochalgorithms.pdf - L3-Stats.pdf - TP1-Sample-Frejus.zip - TP2-QSQ-Frejus.zip - TP3-Kernel-Frejus.zip
- Prof. Jean-Michel Marin (Université de Montpellier, France) - Computational methods for Bayesian inference - etics2019jmm-part1.pdf - etics2019jmm-part2.pdf - etics2019jmm-part3.pdf - etics2019jmm-part4.pdf - etics2019jmm-modelin.pdf - etics2019jmm-capture.pdf - etics2019jmm-exos.zip
- Prof. Youssef Marzouk (Massachusetts Institute of Technology, Cambridge, MA, USA) - Transport methods in Bayesian computation - etics-ymarz-part1.pdf - etics-ymarz-part2.pdf
ETICS2018, June, 4-8, Roscoff, France
Organizers : Jean Giorla (CEA/DAM), Bertrand Iooss (EDF R&D) and Guillaume Perrin (CEA/DAM)
Location: Station Biologique Roscoff, Centre UPMC-CNRS - link
Scientific organizers : Bertrand Iooss (EDF R&D), Jean Giorla and Guillaume Perrin (CEA/DAM)
Lecturers:
- Dr Andrei Bursuc (SafranTech, France) - Deep Learning, a journey from feature extraction and engineering to end-to-end pipelines - ETICS2018_bursuc1.pdf - ETICS2018_bursuc2.7z - ETICS2018_bursuc3a.pdf - ETICS2018_bursuc3b.pdf - ETICS2018_bursuc4.7z - ETICS2018_bursuc5.pdf /* ETICS2018_bursuc1aa.pdf - ETICS2018_bursuc1bb.pdf - ETICS2018_bursuc2.pdf - ETICS2018_bursuc3.zip - ETICS2018_bursuc4.zip */
- Dr Nicolas Gayton (Sigma Clermont, france) - Concepts and methods for robust and/or reliable design - ETICS2018_gayton.pdf
- Prof Sankaran Mahadevan (Vanderbilt University, USA) - Recent developments in Uncertainty Aggregation (forward problem) and Uncertainty Reduction (inverse problem) - ETICS2018_mahadevan.pdf
- Dr Fabrizio Ruggeri (CNR, IMATI, Italy) - Introduction to Robust Bayesian Analysis - ETICS2018_ruggieri.pdf
- Dr Merlin Keller and Jérôme Stenger (EDF R&D, France) - Practical session on Robust Inference using the Python Mystic framework for global optimization - 0_prerequis_tp.pdf - 1_OUQ_intro.pdf - 2_case_study_presentation.pdf - 3_Simulated Annealing_Intro.pdf - 4_Differential_Evolution_Intro.pdf - 5_Practical_Session_Presentation.pdf - 6_Fiche_TP.pdf - 1D.zip - 4D_Correction.zip
- Dr Guillaume Perrin (CEA, France) - Guarantee by simulation - ETICS2018_perrin.pdf
- Dr Mathieu Couplet (EDF R&D, France) - Validation of scientific computing tools - ETICS2018_couplet.pdf
/* Registration - Fees = 800€ (including accomodation, meal, participation to the summer school) */
ETICS2017, October, 1-6, Porquerolles, France
Organizers : Bertrand Iooss (EDF R&D) and Jean Giorla (CEA/DAM)
Lecturers:
- Max Gunzburger (Florida State University, USA): Multifidelity approaches to UQ and Optimal Multilevel Multifidelity Monte Carlo methods - etics2017_gunzburger.pdf
- Luc Pronzato (CNRS - Sophia Antipolis, Nice, France): Optimal Design of Experiments - etics2017_pronzato1.pdf - etics2017_pronzato2.pdf - etics2017_pronzato3.pdf
- Pierre-Henri Wuillemin (Université Pierre et Marie Curie, Paris, France): Applications of probabilistic graphical models - etics2017_wuillemin.pdf
- Sébastien Destercke (Université de Technologie de Compiègne, France): Imprecise probabilities: why, when, how? - etics2017_destercke.pdf
- Nicolas Bousquet (EDF R&D, Chatou, France): Stochastic prior modelling for uncertain inputs - etics2017_bousquet.pdf
ETICS2016, June, 6-10, Barcelonette, France
Scientific organizers : Bertrand Iooss (EDF R&D) and Jean Giorla (CEA/DAM)
Lecturers:
- François Bachoc (Université Paul Sabatier): Calibration of computer experiments
- Sébastien Da Veiga (SafranTech): Advanced methods in sensitivity analysis
- Stéphane Gaïffas (Ecole Polytechnique): Methods for covariance matrix estimation -
- Tim Sullivan (Free University of Berlin / Zuse Institute Berlin): Optimal distributionally robust uncertainty quantification
