Annonce en lien avec l’Action/le Réseau : MACLEAN
Laboratoire/Entreprise : IMT Atlantique, UMR CNRS Lab-STICC
Durée : 36 months
Contact : ronan.fablet@imt-atlantique.fr
Date limite de publication : 2020-03-31
Contexte :
New CNES PhD position for fall 2020 on Deep Learning for Space Oceanography in the framework of AI Chair Oceanix (https://rfablet.github.io/projects/2019-oceanix).
Sujet :
Artificial Intelligence (AI) technologies, models and strategies open new paradigms to address the modeling, simulation, forecasting and reconstruction of complex systems, including ocean-atmosphere dynamics. Due to the irregular space-time sampling of in situ and spaceborne observation data, most envisioned AI-driven strategies rely on learning representations from simulation data and applying these representations to observation data to inform the processes of interest. The applicability of such schemes may then strongly rely on the ability of simulation data to truly match observation data features, which may be questioned for numerous processes. The general objective of this PhD is to investigate the extent to which one may develop fully observation-driven schemes in the context of the space-based observation of flying and future satellite missions such as SWOT mission. From a methodological point of view, we propose to state this challenge for given geophysical processes or variables as a joint end-to-end learning of a latent geophysically-sound representation and of the associated inversion scheme from irregularly-sampled observation dataset. Using CNNs (Convolutional Neural Network), the taregted methodological contributions are regarded as key building blocks to revisit earth observation challenges, including among others operational satellite-derived geophysical products, data-driven schemes for inter-comparison studies between observation and/or simulation data,… This PhD will be implemented in the collaborative framework of Melody project (ANR MN 2020-2022) with strong interactions between Lab-STICC (R. Fablet), LOPS (B. Chapron), IGE/MEOM (J. Le Sommer), OceanNext (C. Ubelman) and OceanDataLab (L. Gaultier). SWOT-related case-studies in Melody, e.g. wave-current separation, SWOT-derived SLA L4 products, will be the core application ground for the considered methodological developments, including the exploitation of SWOT fast-sampling phase data. The PhD candidate will benefit from the gathered multidisciplinary expertise of the supervision team in Ocean Science, Ocean Remote Sensing, Fluid Dynamics and Data Science.
Detailed presentation of the PhD: https://rfablet.github.io/files/phd_proposal_rfablet_CNESMelody_201910_1.pdf
Profil du candidat :
The targeted PhD candidate shall have a MSc and/or engineer degree in Data Science or Artificial Intelligence with a strong interest in environmental sciences, possibly acknowledged by previous activities or experience. A dual degree in ocean science and data science as promoted by Isblue MSc program would be of key interest. Besides a strong theoretical background, computer skills, including first experience in using state-of-the-art deep learning frameworks (e.g., tensorflow, pytorch) and programming environment (e.g., python, git server), will be expected.
Formation et compétences requises :
More information on the application procedure at: https://recrutement.cnes.fr/en/annonce/899896-200-end-to-end-learning-of-geophysically-sound-cnn-representations-29200-brest
Adresse d’emploi :
IMT Atlantique, Brest, France