Workshop on AI for Geophysical Dynamics
November 12-15 2018, Brest
Objectives : the objectives of this collaborative workshop on Artificial Intelligence and Machine Learning for Geophysical Dynamics are three-fold:
- Review, discuss & advance machine learning and deep learning for the analysis, modeling, forecasting and reconstruction of ocean dynamics
- Stimulate the cross-fertilization between AI & Ocean science
- Bridge model-driven and data-driven frameworks
- Initiate the definition and evaluation of proof-of-concepts for case-studies of interest
Invited speakers : (see keynote titles and abstrats)
- Prof. Nicolas Courty, Univ. Bretagne Sud/IRISA, France (webpage)
- Dr. Philippe Naveau, CNRS/IPSL, Francce (webpage)
- Dr. Julien Le Sommer, CNRS/IGE, France (webpage)
- Prof. François Lauze, DIKU, Denmark (webpage)
- Prof. François Rousseau, IMT Atlantique, France
- Prof. Steve Brunton, Univ. Washington, USA (webpage)
Organizing committee :
- Prof. Ronan Fablet, IMT Atlantique/Lab-STICC, France
- Prof. François Rousseau, IMT Atlantique/Latim, France
- Prof. Lucas Drumetz, IMT Atlantique/Lab-STICC, France
- Dr. Lionel Mathelin, CNRS/LIMSI, France
- Dr. Bérengère Podvin, CNRS/LIMSI, Franc
- Prof. Eric Blayo, UGA/INRIA AirSea, France
- Prof. Maëlle Nodet, UGA/INRIA AirSea, France
Location: Pôle Numérique Brest Iroise, Technopôle Brest-Iroise, 29280 Plouzané (google map link).
Preliminary program: The workshop will combine scientific talks and open forum sessions as detailed below.
Sponsors: The workshop is supported by IMT Atlantique, Labex Cominlabs (project SEACS), Microsoft AI EU Ocean award (AI4OceanDyn).
| Monday, November 12
2.00pm-6.00pm |
Welcome word
Keynote: Dr. J. Le Sommer, CNRS, IGE (45’+15’) Ocean Science & Machine Learning: Challenges & Opportunities
Keynote: Prof. F. Rousseau, IMT Atlantique (45’+15’) Registration, Diffeomorphism, ODE & ResNet
Open forum session (Propose & share problem/theme proposal) 5′ presentation for proposals of working group themes/subjects (any participant) |
| Tuesday, November 13
9.00am-12.30pm |
Keynote: Prof. N. Courty, Univ. Bretagne Sud, IRISA (45’+15’)
Machine learning & Optimal transport
In parallel : Introduction to Machine learning and Deep learning for beginners Open forum session & working groups |
| Tuesday, November 13
2.00pm-6.00pm |
Keynote: Prof. S. Brunton, Univ. of Washington (45’+15’)
Machine learning of interpretable nonlinear models for unsteady flow physics
Open forum session & working groups |
| Wedsneday, Nov. 14
9.00pm-12.30pm |
Keynote: Dr. P. Naveau, CNRS, IPSL (45’+15’)
Learning complex spatial dependence structure among heavy rainfall or how to combine data within max-stable processes? (see abstrat)
Open forum session & working groups |
| Wedsneday, Nov. 14
2.00pm-6.00pm |
Keynote: Prof. F.. Lauze, DIKU (45’+15’)
To be defined
Open forum session & working groups |
| Thursday, November 15
9.00pm-12.00pm |
Wrap-up session |
List of participants:
- Cédric Herzet, SIm-SMART, INRIA Rennes
- Etienne Mémin, Fluminance, INRIA Rennes
- Lucas Drumetz, IMT Atlantique, Lab-STICC
- Pierre Tandeo, IMT Atlantique, Lab-STICC
- François Rousseau, IMT Atlantique, Latim
- Ronan Fablet, IMT Atlantique, Lab-STICC
- Said Ouala, IMT Atlantique, Lab-STICC
- Duong Nguyen, IMT Atlantique, Lab-STICC
- Manuel Lopez Radcenco, IMT Atlantique, Lab-STICC
- Yicun Zheng, IMT Atlantique, Lab-STICC
- Paul Platzer, IMT Atlantique, Lab-STICC
- Dorian Cazau, IMT Atlantique, Lab-STICC
- Julien Le Sommer, CNRS, IGE
- Redouane Lguensat, CNRS, IGE
- Nicolas Courty, UBS, IRISA
- Clément Le Goff, E-odyn
- Béregngère Podvin, CNRS, LIMSI
- Eric Blayo, UGA, LJK
- Steven Brunton, University of Washington
- Bertrand Chapron, Ifremer, LOPS
- François Lauze, DIKU
- Elodie Martinez, IRD, LOPS
- Yvain Quéau, CNRS, Greyc
- Marie-Noëlle Bouin, CNRM
- Pierre Ailliot, UBO, LMBA
- Thomas Corpetti, CNRS, LETG
- Nicolas Longépé, CLS
- Sammy Metref, IGE
- Jacques Verron, IGE
- Alex Ayet, Ifremer/LOPS, ENS/LMD
- Elise Arnaud, AirSea, LJK
- Arthur Vidard, Airsea, LJK
- Jean-Marc Delouis, LOPS
- Laurent Cordier, CNRS, PRIME
- Gilles Tissot, Fluminance, INRIA Rennes
- Fabrice Collard, ODL
Below a first list of potential themes of interest for the working groups :
- Learning-based identification of geophysically-sound governing equations (R. Fablet)
- Learning Poincaré-Takens-Koopman operators (B. Chapron)
- Learning & Extremes of dynamical systems (P. Naveau)
- NN-based Kalman-like assimilation (L. Drumetz)
- Diffeomorphic flows, Dynamical Systems & Deep Learning (F. Rousseau)
- Learning closure representations (J. Le Sommer)
- Bulk formulation & Machine Learning (E. Blayo/M.N. Bouin)
- Learning-based modeling and denoising of satellite attitude dynamcis (F. Collard)
