Florent Sureau

---

Research Interests

Since 2009 in Cosmostat, my main interests are on applying various advanced optimization techniques to solve inverse problems, with application in astrophysics. This implies adapting algorithms for the specific application and task considered, as well as finding efficient ways of applying them to process generally large volume of data, whether for CMB or for weak lensing data analysis.

I have contributed to Planck data analysis using sparsity, developing tools for (polarised) data defined on the sphere, with Jérome Bobin and Jean-Luc Starck.  My main projects were to develop  component restoration using constrained convex optimization, and compact source removal in Cosmic Microwave Background (CMB) datasets.  I was part of the Planck consortium which has received  the 2018 Gruber Price in Cosmology.

On weak lensing data analysis, I  have contributed to the shear measurement pipeline developed between CEA/Cosmostat and EPFL (M. Gentile,  F. Courbin) that finished third on the GREAT3 Challenge (http://great3challenge.info).  I investigated in particular several improvements and strategies to include sparsity in forward-fitting shape measurement techniques.  I am also involved in the LENA ERC project of Jérome Bobin where I contribute to better understand how galaxy morphological parameters impact shape and shear measurement (with Arnau Pujol,  Jérome Bobin, Martin Kilbinger). I am also part of the Euclid Consortium, responsible for coordinating LE3 data models (see http://www.cosmostat.org/projects/euclid ) and defining the data models for weak-lensing algorithms.  

Finally, I contributed to the FET-OPEN Dedale project  (Sept. 2015- Sept. 2018) leaded by  Jean-Luc Starck, where I was responsible for the work-package dedicated on signal processing methods, and mainly interested by learning applications.  I developed  in particular dictionary learning methods for data on the circle that can be applied to polarised data, and worked on extending patch-based dictionary learning methods for data defined on the sphere (choosing the HEALPix  framework to form charts on the sphere). In that project, I also proposed dictionary learning methods for spectroscopic and photometric redshift estimation, and contributed to develop unsupervised feature learning techniques for galaxy SED representation and for spectroscopic redshift estimation  with Joana Frontera-Pons and Jérome Bobin.

Teaching Activities

• Practical exercices for the Ecole Doctorale Astronomie et Astrophysique d’Ile de France (2011-2013): half-day exercises on applying sparse representations to solve inverse problems in astrophysics
• Teaching and Practical exercices for the Ecole Doctorale Astronomie et Astrophysique d’Ile de France ( 2013-2016): a day of course and exercises on applying sparse representations to solve inverse problems in astrophysics
• Teaching in master IMA UPMC/Telecom-ParisTech (2016-2018): half-day of courses on sparse representations and inverse problems
• Tutorials given at the ADA VII summer school Advanced data processing and reproducible research: two 3-hour presentations and exercises on Deconvolution and Compressed Sensing and Data Analysis on the sphere 
• Course on Image Processing set up with Pr. Nguyen-Verger (Universite de Cergy Pontoise, 2007): 25 hours of course and practical exercices for Licence 3 Maths and computer science students

Education/Past Experiences

I obtained my Ph.D. of Physics from the Universite Paris XI in 2008, for my thesis on reconstruction techniques for Positron Emission Tomography, under the supervision of Régine Trébossen, Andrew Reader and Bertrand Tavitian at the Service Hospitalier Frédéric Joliot (CEA). I then worked for one year in Michel Defrise's lab at the Vrije Universiteit Brussels, working in particular on discretization strategies to implement limited data reconstruction in computerized tomography. Prior to this, I graduated in 2004 from the University of Virginia (M.Sc.) and from the Ecole Centrale de Lille (Diplôme d'ingénieur).

---