Stellenangebot

PhD position in University of Paris-Saclay, Paris, France

Summary:
We are looking to fill a PhD position for the project of “Defining domain of validity in AI (Artificial Intelligence) with conformal prediction: Application to high frequency time series”.

Location: University of Paris-Saclay, Paris, France
Application deadline: 28 April 2022
Keywords: statistics, machine learning, data science, uncertainty quantification, functional data

Project description:

The construction of a trusted AI is one of the major challenges of AI to enable its adoption. Trust is at the heart of the debates (AI regulatory project at European level) and of the concerns of manufacturers.

To identify the area of ​​reliable operation of an AI, we propose conformal prediction techniques. These "Distribution-free" techniques are particularly suitable for calculations of prediction intervals of complex Machine Learning and Deep Learning models, and provide the means to map the uncertainty of AI.

The objective of the thesis is to contribute to a formalization and a precise quantification of the notion of domain of validity thanks to the Quantification of Uncertainties, including in the case of variables of very large dimensions, such as time series and functional data.

As part of the project, open source applications and development (#MAPIE https://github.com/scikit-learn-contrib/MAPIE) are planned.

https://opengraph.githubassets.com/7ed8591de32c1ae952c06dd7dd26e845b35fe27434d28cc986e1e534f9e3c3f3/scikit-learn-contrib/MAPIE

GitHub - scikit-learn-contrib/MAPIE: A scikit-learn-compatible module for estimating prediction intervals.
github.com
A scikit-learn-compatible module for estimating prediction intervals. - GitHub - scikit-learn-contrib/MAPIE: A scikit-learn-compatible module for estimating prediction intervals.

The subject is accessible in detail at the following address on the Doctoral School website:
https://www.adum.fr/as/ed/voirproposition.pl?langue=&site=psedmh&matricule_prop=41206

Practical Information:
The thesis will take place at the Laboratory of Mathematics and Modeling of Evry (LaMME) and benefits from UDOPIA funding (https://www.universite-paris-saclay.fr/en/udopia-doctoral-program-artificial-intelligence), the doctoral program in Artificial Intelligence from Paris Saclay University.

UDOPIA doctoral program in artificial intelligence | Université Paris-Saclay
www.universite-paris-saclay.fr

Entry Requirements:
We are looking for a candidate who has (or is about to) hold a Master's degree in Mathematics, Statistics/Machine Learning, Data Science.

Those interested should contact supervisors: Juhyun Park (juhyun.park@ensiie.fr) & Nicolas Brunel (nicolas.brunel@ensiie.fr)

Zurück zur Übersicht