MASCOT PhD student 2020 Meeting

Due to the cancellation of the MASCOT-NUM 2020 meeting, a special meeting devoted to the MASCOT-NUM PhD students' day is organized from September 17th to September 18th, 2020, in Grenoble Alpes University (France).

It will be located at the Auditorium of the IMAG building, Grenoble University.

This meeting is devoted to PhD students communications on topics related to the Mascot-Num network activities, i.e. sensitivity analysis, uncertainty quantification in simulation, design and modelling of computer experiments, model calibration and validation, structural reliability, optimization under uncertainty, data-driven modelling methods, etc.

For the poster session, PhD students are invited to submit a two-page abstract using the LaTeX abstract template to be held on the same day. The PDF file shall be sent by May 11th, 2020 (deadline extended!) to

A prize of 1000€ will be awarded for the best oral presentation by the Mascot-Num scientific committee. A prize of 500€ will be awarded for the best poster by the Mascot-Num scientific committee.

All the speech will be in English.

The best student presentation award has been granted to Jérôme Stenger (EDF R&D/Université de Toulouse).
The best student poster award has been granted to Clément Gauchy (CEA/Ecole Polytechnique).

Program & Handouts

Day 1: September 17th

11h45 Welcome buffet

13h00-13h15 Welcome speech

13h15-14h15 Introductive lecture: Romain Couillet (CentraleSupelec; University ParisSaclay; IDEX GSTATS Chair & MIAI LargeDATA Chair, University Grenoble-Alpes) - Can random matrices change the future of machine learning? - couillet2020slides.pdf

14h15-14h45 Poster blitz 1: Antoine Ajenjo, Bruno Barracosa, Clément Bénard, Alexis Cousin, Jhouben Cuesta, Nestor Demeure, Clément Gauchy, Thierry Gonon

14h45-15h15 Coffee break

15h15-15h45 PhD 1: Jérôme Stenger (Univ. Toulouse, EDF R&D) - Optimal Uncertainty Quantification of a Risk Measurement - stenger2020.pdf - stenger2020slides.pdf

15h45-16h15 PhD 2: (visioconference) Deyu Ming (Univ. College London) - Integrated Emulators for Systems of Computer Models - ming2020.pdf - ming2020slides.pdf

16h15-16h45 Poster blitz 2: Adrien Hirvoas, Corentin Houpert, Baptiste Kerleguer, Henri Mermoz Kouye, Sébastien Petit, Alvaro Rollon, Martin Wieskotten

16h45-18h15 Poster session

Day 2: September 18th

9h30-10h00 PhD 3: Cécile Haberstich (Ecole Centrale Nantes, CEA/DAM) - Principal Component Analysis and boosted optimal weighted least-squares for training tree tensor networks - haberstich2020.pdf - haberstich2020slides.pdf

10h00-10h30 PhD 4: Maria-Belen Heredia (INRAE, Univ. Grenoble Alpes) - Aggregated Shapley effects from an acceptance-rejection sample: application to an avalanche model - heredia2020.pdf - heredia2020slides.pdf

10h30-11h PhD 5: (visioconference) Louise Kimpton (Univ. of Exeter) - Correlated Bernoulli Processes Using De Bruijn Graphs - kimpton2020.pdf - kimpton2020slides.pdf

11h-11h30 Coffee break

11h30-12h00 PhD 6: Thomas Bittar (Ecole des Ponts ParisTech, EDF R&D) - A Decomposition Method by Interaction Prediction for the Optimization of Maintenance Scheduling - bittar2020.pdf - bittar2020slides.pdf

12h00-12h30 PhD 7: Victor Trappler (Univ. Grenoble-Alpes) - Robust Calibration of numerical models based on relative regret - trappler2020.pdf - trappler2020slides.pdf

12h30-14h00 Buffet (+ meeting of the jury for the attribution of the prices)

14h00 Attribution of the prices

14h30 End

Welcome speech

Speaker: Romain COUILLET (professor at CentraleSupélec, University ParisSaclay; IDEX GSTATS Chair & MIAI LargeDATA Chair, University Grenoble-Alpes)

Title: Can random matrices change the future of machine learning?

Abstract: Many standard machine learning algorithms and intuitions are known to misbehave, if not dramatically collapse, when operated on large dimensional data. In this talk, we will show that large dimensional statistics, and particularly random matrix theory, not only can elucidate this behavior but provides a new set of tools to understand and (sometimes drastically) improve machine learning algorithms. Besides, we will show that our various theoretical findings are provably applicable to very realistic and not-so-large dimensional data.

List of posters