AMI « Apport du Numérique au domaine du recyclage »
Le PEPR « Recyclage » lancé au printemps 2023 est structuré en : 5 axes verticaux « matériaux » (plastiques,...
24 Octobre 2024
Catégorie : Stagiaire
Unrolled proximal algorithms for estimating the reproduction number of Covid-19 pandemic
Master internship of 4 to 6 months in 2025 located at Laboratoire des Sciences du Numérique de Nantes (LS2N).
Supervision and contact: Barbara Pascal (barbara.pascal@cnrs.fr) and Sébastien Bourguignon (sebastien.bourguignon@ec-nantes.fr).
Application: Send a CV, master grades, references and motivations to B. Pascal and S. Bourguignon.
Epidemics, striking heavily our societies both on the sanitary, economic and societal sides, are a major public health issue. The crisis triggered by the Covid-19 pandemic emphasized the need for accurate, robust and real-time tools for assessing epidemic intensity.
A key epidemic indicator is the effective reproduction number of an epidemic, defined as the time-varying average number of secondary infections caused by one standard contagious individual: when it is larger than one, this indicates a phase of exponential growth of the number of new infections, at risk to induce a severe epidemic wave. A major challenge in estimating the reproduction number in real time during a pandemic crisis is that the available data are of low quality, calling for the development of robust estimators. To that aim, penalized likelihood estimators leveraging the variational framework have recently been proposed, involving proximal algorithms.
The main objective of this internship is to extend this fully model-based approach into an hybrid physically informed deep learning estimating procedure, using state-of-the-art unrolling strategies.
More informations are available at: https://bpascal-fr.github.io/assets/pdfs/internship2425_unroll-covid_LS2N.pdf and contacting Barbara Pascal (barbara.pascal@cnrs.fr).