Uncertainty quantification: theory and application
From Wednesday 2 May 2018 to Thursday 3 May 2018
Training
Announced
Deadline for registration: 15 days before the starting date of each training
Duration : 2 days / (14 hours)
Abstract
This training course is given by three specialists of uncertainty quantification: Dr Pietro Marco Congedo – researcher at INRIA Bordeaux Sud-Ouest, Dr Vivien Mallet – researcher at INRIA Paris Rocquencourt, and Dr Pierre Sagaut – professor at Laboratory M2P2, Aix-Marseille University. The first half of the course focuses on theoretical aspects: uncertainties definition, Monte Carlo methods, projection methods, approximation methods and calibration data methods. The second half of the course is dedicated to application of uncertainty quantification to the fields of algorithm, CFD and geosciences.
Target participants
This training session is for engineers, physicists, computer scientists
and numerical analysts who wish to learn about uncertainty quantification.
Prerequisites
None.
Scientific contacts :
– Dr Pietro Marco Congedo, INRIA Bordeaux Sud-Ouest
– Dr Vivien Mallet, INRIA Paris Rocquencourt
– Pr Pierre Sagaut, Laboratory M2P2 – Aix-Marseille University
Fee
- Trainees/PhDs/PostDocs : 120 €
- CERFACS shareholders/CNRS/INRIA : 360 €
- Public : 720 €
Program
(Everyday from 9h00 to 17h30)
Day1 Theoretical aspects (Pr P. Sagaut)
- Introduction and definition: errors, uncertainty, risk, stochastic modelling, sources of uncertainty
- Monte Carlo method and variants Projection methods: Galerkin methods and collocation, generalized polynomial chaos, implantation, etc.
- Methods of approximation: kriging, co-kriging, prediction error, POD
- Calibration data: Bayesian inference
Day 2 morning: Application to algorithms (Dr P.M. Congedo)
- Innovative algorithm for quantifying uncertainties in CFD applications
- Non-probabilistic approach and epistemic uncertainty
- Optimisation under uncertainties
Day 2 afternoon: Application to geosciences (Dr V. Mallet)
- Ensemble evaluation, Data assimilation
- The multiple models
- Ensemble evaluation
- A posteriori uncertainty estimate for data assimilation
