Keywords: Machine Learning, Graph Mining, Representation Learning, Transfer Learning

Description: Representation learning for graph mining is a significant challenge. Data in the form of graphs has become ubiquitous for describing complex information or structures. In the social sciences, graphs will make it possible to study relationships and interactions between people; in biology, we will focus on graphs to model genetic interactions or metabolic networks. However, compared to images or text, the structure of a graph is very irregular making learning a good representation more difficult. This post-doctorate is articulated around 2 research axes. The first direction is to look at representation transfer for graph mining. In this context, the objective will be to work on learning representations which are, on the one hand, specific to the tasks that one wishes to solve, for example the prediction of link, the classification of nodes or the detection of communities, but also exploit the potential dependencies between these different tasks. This first axis naturally opens the way to the processing of dynamic graphs. In this case, the graphs are most often represented as a collection of static graphs at different times, and for each graph a new representation is systematically learned. The objective is to develop a learning framework capable of detecting significant changes in the structure of the same graph, and to adapt the representation learned so as to transfer the immutable characteristics and thus to preserve the knowledge already acquired.

Required profile: The candidate has a Ph.D. and will have a strong background in machine learning including a good foundation in statistical learning and mathematics. He is also expected to have a good level in programming. Finally, the candidate must also have a good level of English and have both an interest in theoretical and practical aspects.

Supervisors:
Charlotte Laclau, charlotte.laclau@univ-st-etienne.fr
Baptiste Jeudy, baptiste.jeudy@univ-st-etienne.fr

Application: The starting date is flexible, between February and September 2019.The candidate should indicate the preferred starting date in the application email.
Applications including a CV, a list of publications and an approximately two-page description of research interests should be sent by email to both supervisors’ emails. Applicants should also arrange at least one recommendation letter (to be sent to the same address). Informal inquiries can be sent to the same address

Les candidats sont fortement encouragés à prendre contact avec les équipes du département afin d’affiner leur projet et de préciser les thématiques et les projets d’intérêt.
Date limite de réception des candidatures : 28 Février 2019

Pour complément d’information :
– Emmanuel Rachelson, emmanuel.rachelson@isae-supaero.fr
– Carlos Aguilar Melchor, carlos.aguilar-melchor@isae-supaero.fr

Cordialement,
Emmanuel

To encourage mobility, the EU mandates that our recruits have not spent more than 12 months in the lab’s country over the 3 years prior to the recruitment date.

For details and application, go to https://euraxess.ec.europa.eu/jobs/363030 This first call closes on January 15, 2019.