|
|
Nuts second workshop Welcome to the Réseau Numérique en Terre Solide (NuTS), a CNRS thematic network (RT) that aims to foster the community of developers and users of numerical tools for solid earth science. Our goal is to make this community more visible, efficient, and informed through a series of conferences and training courses on digital science, from data manipulation to high-performance scientific computing. We are excited to announce the second workshop of the RT NuTS, which will take place from May 28th to May 31st in Strasbourg. This workshop is dedicated to "inverse problems" in solid earth science with a focus on numerical and computational aspects. It is divided into two parts: the first two days will feature scientific talks, and the last two days will offer hands-on practicals on two topics (see the program below). During the practicals, attendees will have the opportunity to work with Jupyter Notebooks and the Python language, using the cutting-edge computing resources provided by the GLiCID mesocentre, including a series of A100 GPU. The practicals will cover various topics, from Bayesian inference and uncertainty quantification hands-on to automatic differentiation in the Julia language. We are planning a poster session dedicated to inverse problem-related studies, especially those that are in progress and require further discussion. We welcome contributions from students and researchers interested in sharing their plans and unfinished work with the community. We invite you to join us for this exciting workshop, where you will learn from experts in the field, gain hands-on experience with inverse problem-related practical programming, and network with other researchers in solid earth science. Do not miss this opportunity to advance your skills and knowledge in numerical and computational science. To register for the workshop, please click on the "Register Now" button. If you have any questions or need assistance with registration, please don't hesitate to contact us. We look forward to seeing you in Strasbourg! Event location:
The workshop will take place at the "EOST auspices - Manufacture des Tabacs", 7 Rue de la Krutenau, 67000 Strasbourg
Program Preliminary Program The program is still under construction Tuesday, May 28th (Afternoon) Introduction: Yann Capdeville Poster pitches: concise oral presentation of the posters Special: Emmanuel Chaljub (ISTerre). Title: ¨FormaTerre in Data Terra: contribution of the Solid Earth community to the E-infrastructure of the Earth & Environment System. Plenary Lectures:
Wednesday, May 29th (all day) Confirmed speakers:
Thursday, May 30th (all day) Morning: Practicals: Bayesian inference with Markov chain Monte Carlo sampling Thomas Bodin (LGLTPE) Description: This short course will introduce Bayesian inference and uncertainty quantification. We will see how a probabilistic framework can be used to tackle highly non-linear and ill-posed inverse problems, and how data errors can be propagated towards model uncertainties. Moving beyond theory, the lab component of the course will introduce participants to Markov Chain Monte Carlo methods, a computational tool for approximating probabilistic solutions in geophysical inverse problems. Participants will gain practical experience by coding this algorithm from scratch in a simple toy problem. Afternoon: Practicals: High-performance computing in geosciences and adjoint-based optimisation with Julia on GPUs (part 1)
Ludovic Räss (ETH Zurich) and Ivan Utkin (ETH Zurich)
This workshop offers hands-on experience in high-performance computing for geosciences, focusing on solving and constraining partial differential equations (PDE) using GPUs. The goal of the workshop is to develop a fast iterative GPU-based solver for elliptic equations and use it to:
1. Solve a steady state subsurface flow problem (geothermal operations, injection and extraction of fluids) 2. Invert for the subsurface permeability having a sparse array of fluid pressure observations We will not use any "black-box" tooling but rather develop concise and performant codes (300 lines of code, max) that execute on GPUs. We will also use automatic differentiation (AD) capabilities and the differentiable Julia stack to automatise the calculation of the adjoint solutions in the gradient-based inversion procedure.
Friday, May 31th (all day) All Day Practicals: High-performance computing in geosciences and adjoint-based optimisation with Julia on GPUs (part 2) |
Online user: 2 | Privacy |