Accueil > Actualités > Séminaires > Séminaire de Redouane Lguensat


Titre : Inferring hidden equations using Quasi-Geostrophic theory guided machine learning
Nom du conférencier : Redouane Lguensat
Son affiliation : Université Grenoble Alpes, IGE, Grenoble
Laboratoire organisateur : IPSL
Date et heure : 20-03-2019 11h00
Lieu : Campus Pierre et Marie Curie - LATMOS, grande salle du 4ème, Tour 45-46, pièce 411
Résumé :

Inferring  hidden  equations  governing  dynamical  systems  from data  has  always  been  one  of  the  challenging problems  in  the interplay  between  physics  and  data  science. It  was  just  a matter  of  time  before  the  recent advancements in machine learning and in computational capacities come in hand and spark off a series of works dedicated to address this problem.

In this  work  we  present  a  Quasi-Geostrophic  numerical  model coded  using  differentiable  operators  thus permitting  the  use  of  automatic  differentiation  libraries  (e.g.  Tensorflow).  This makes  the  model  flexible  and suited for parameter optimization, especially using neural networks. We illustrate the relevance of the proposed architecture through an example of a regression problem where we show how can we obtain the parameters of the potential vorticity equation using only consecutive scenes of Sea Surface Height. This can be of interest for finding the closest QG-like approximation to a given ocean simulation model or help exploring the effect of adding new operators in the potential vorticity equation. The code we provide is suitable for GPU implementation and therefore can allow for faster execution and profit from the quick advancements in GPU development.

We  expect  that  the  directions  of  research  we  suggest  will help  in  bringing  more  interest  in  applied machine learning to ocean numerical modeling.

Contact :