Bartlett, Chiarenza, Doeser, Leclercq
Prideaux-Ghee, Leclercq, Lavaux, Heavens, Jasche
Leclercq, Heavens
Leclercq, Faure, Lavaux, Wandelt, Jaffe, Heavens, Percival, Noûs
Lavaux, Jasche, Leclercq
Leclercq, Enzi, Jasche, Heavens
Leclercq
My name is Florent Leclercq. I am a research scientist (Chargé de recherche CNRS) at the Institut d'Astrophysique de Paris (IAP). I hold an interdisciplinary position at the interface between astrophysics (Institut national des sciences de l'Univers, INSU) and information science (Institut des sciences de l'information et de leurs interactions, INS2I). I work in the fields of numerical cosmology and artificial intelligence, focusing in particular on the analysis of galaxy survey data. I have been a member of the Aquila Consortium since it was created, in 2016. I am also a member of the Euclid Consortium, where I currently co-lead the "Additional Probes" work package of the Galaxy Clustering Science Working Group.
My current research interests are related to the study of the cosmological large-scale structure using statistical inference, machine learning, and high-performance computing tools. I am particularly interested constraining cosmology from the large-scale structure, in the initial conditions from which it originates, its formation history and the description of the cosmic web.
Please check the full version of my CV, my list of publications and my list of communications.
5/5 We're excited to work on these developments in the context of next-generation surveys, in particular @ec_euclid.
Developed in @AquilaScience at @astroIAP with Mayeul Aubin (IAP), Deaglan Bartlett (IAP), Marco Chiarenza (IAP, U. Milan), Ludvig Doeser (Stockholm University), Tristan Hoellinger (IAP), and Guilhem Lavaux (IAP).
4/5 spatial COmoving Lagrangian Acceleration (sCOLA, Leclercq et al. 2020, arxiv.org/abs/2003.04925) uses a large-scale analytical solution to spatially split simulations in independent tiles. It achieves perfect parallelism by fully removing the need for communications across the full computational volume. Also, it allows for fast lightcone and mock catalogue generation (work in progress).
3/5 (temporal) COmoving Computer Acceleration (tCOCA, Bartlett et al. 2024, arxiv.org/abs/2409.02154) reimagines the use of neural networks for emulating N-body simulations. It generalises the idea of tCOLA: running simulations in a new frame of reference. But it is not an emulator! It solves the correct equations of motion, so it is an ML-safe use of neural networks. Explainability is not needed! COCA makes simulations cheaper by skipping unnecessary force evaluations.
2/5 I centered my talk around the concepts of (i) ML-safety and (ii) perfect parallelism. I call a technique "ML-safe" if it ensures that the results are either correct by construction or, at worst, suboptimal. And a "perfectly parallel" computational task is one made of independent sub-tasks with no communication between them, allowing for highly efficient parallel processing. These two concepts are central to our recent progress on efficiently simulating gravity.
In this post, we discuss a new algorithm to simulate the Universe on a variety of hardware architectures.
I recently scanned the literature for the purpose of following and plotting the number of particles used in \(N\)-body simulations over the last five decades.
08-04-2020
Likelihood-free inference provides a framework for performing Bayesian inference in cosmology, by replacing likelihood calculations with data model evaluations.
25-04-2019
pySELFI is a publicly-available implementation of the Simulator Expansion for Likelihood-Free Inference algorithm, allowing primordial power spectrum inference from black-box galaxy surveys.
Simbelmynë is a publicly-available simulator to generate synthetic galaxy survey data. A more detailed description of the code can be found on its homepage, hosted on this website.
This website hosts the BORG SDSS data release, a set of data products that follow a chrono-cosmographic analysis of the three-dimensional large-scale structure of the nearby Universe.