Annonce en lien avec l’Action/le Réseau : aucun

Laboratoire/Entreprise : IMT atlantique, Lab-STICC
Durée : 6 mois
Contact : ronan.fablet@imt-atlantique.fr
Date limite de publication : 2019-12-31

Contexte :
Invertible neural networks (INN) have recently involved a significant research interest. Such representations jointly embed the prediction of the outputs given the inputs as well as the inputs given the outputs. This property is particularly apealing when dealing with inverse problems, i.e. aiming to reconstruct some hidden processes from some observed variables.

In the context of the space-based remote sensing of the oceans, a variety of satellite missions provide observations of sea surface parameters (e.g., temperature, salinity, current). We may now benefit from such large-scale observation datasets to explore, characterize and model upper ocean dynamics. In this respect, theoretical evidence has been provided that sea surface tracers may exhibit relationships, which relate to specific dynamical regimes.

This internship is proposed in the framework of ANR Melody (Bridging geophysics and MachinE Learning for the modeling, simulation and reconstruction of Ocean DYnamics, PI: R. Fablet) and ERC STUOD (Stochastic Transport in Upper Ocean Dynamics, PI: B. Chapron).

Sujet :
In the spirit of our previous work, the goal of this internship will be to explore ocean remote sensing datasets using deep learning strategies to reveal new data-driven representations of upper ocean dynamics. The focus will be given to INN representations.

Profil du candidat :
Msc./Eng. degree in Applied Math., Data Science and/or Physical Oceanography.

Formation et compétences requises :
Good background in data science an applied statistics. Knowledge on deep learning models and experience in deep learning frameowrks (eg, tensorflow, keras, pytorch) would be a plus.

Adresse d’emploi :
IMT Atlantique, technopôle Brest-Iroise, Brest

Document attaché : internship_proposal_MultimodalRepresentation.pdf