Intern on deep learning project : classification of large sets of biological images by convolutional neural networks.

Type de poste
Dates
Durée du poste
Contrat renouvelable
Contrat non renouvelable
Date de prise de fonction
Date de fin de validité de l'annonce
Localisation
Adresse

<p>46 rue d&#39;Ulm 75005</p>
Paris
France

Contacts
Champetier Tiphaine
Email du/des contacts
tiphaine.champetier@ens.fr
auguste.genovesio@ens.fr
Description

Deep learning: classification of large sets of biological images by convolutional neural networks.

Context :

The high-content screening, used in fundamental research and for therapeutic molecules discovery, makes it possible to automatically visualize the effect of a large number of parallel perturbations on a cellular model. This process, which uses automated microscopy, generates large amounts of images. Our laboratory uses and develops deep learning approaches to classify these disturbances automatically and to allow the inference of morphological modifications induced by them.

Goals:

The general goal of this project is to improve the understanding of a cellular model with deep learning.

Expected work:

After a quick upgrade of the candidate on the practical aspects of the application of deep learning methods to large image datasets, the internship project will focus on studying the capacity of networks, pre-trained on natural image classification, to transfer their expertise to biological domain images, and this for classification and clustering problems. We wish to create a transferable system allowing the explanation of biological visual phenotypes. In a second step, it will be possible to train a network with a special loss function in order to obtain a similarity metrics, allowing better clustering results.

Technical skills and education :

This M2 internship may be suitable for people in engineering or computer sciences. It may also be suitable for students in physics or statistics. Knowledge in Applied Mathematics or Statistical Learning is welcome but not absolutely necessary. Applicants will need to have programming basics, ideally in Python and if possible comftable with linux environment. All applications will be studied.