2022-04835 - Post-Doctoral Research Visit F/M HPC for learning in Tree-based tensor format
Le descriptif de l’offre ci-dessous est en Anglais

Type de contrat : CDD

Niveau de diplôme exigé : Thèse ou équivalent

Fonction : Post-Doctorant

Niveau d'expérience souhaité : De 3 à 5 ans

Contexte et atouts du poste

The position is open in the context of the joint project between Airbus R C& T, Cerfacs and Inria.

In the context of supervised learning (e.g., learning regression models as approximation of functions) or unsupervised learning (probability distributions learning), we face CPU time issues as the dimension grows.

We encounter these problems of supervised learning in high dimension when we are interested in the prediction of physical quantities which are very often spatial fields. As an example, we can mention the learning of wall laws in a fluid calculation which allows us not to refine too much near a wall.



Mission confiée

For both continuous graphical models and deep tensor trees, two main operations are particularly costly:

  • Algebraic operations on tensors;
  • The exploration and quantization of many tree or graphical model configurations (Directed Acyclic Graphs).

In the case of tensor networks, learning the coefficients in the leaves of the tree is also expensive (alternate least-squares).



Principales activités

The objective of this post-doc is to accelerate these three functionalities by using different strategies.
Concerning the tensor algebra, we will allow ourselves not to make exact calculations if the improvement is important and the error is controlled.
For the exploration, we can set up strategies of exploration by neighborhoods which implement calculations by batches distributed on the various resources.
Regarding the learning of the coefficients, while the alternate least squares algorithm is a well-stablished method, new algorithms have been developed as well as new tensor formats. It is expected to implement and test these algorithms.

The Celeste library is a C++ library for tensor computation based on the Tucker and Tensor Train formats. Our objective from these applications is to validate and improve the performance of our methods and to integrate it into the netwkork tensor and the graphical models


  • Subsidized meals
  • Partial reimbursement of public transport costs
  • Possibility of teleworking and flexible organization of working hours
  • Professional equipment available (videoconferencing, loan of computer equipment, etc.)
  • Social, cultural and sports events and activities
  • Access to vocational training
  • Social security coverage


2653€ / month (before taxs)