François Lanusse

I am a CNRS researcher working at the intersection between Deep Learning, Statistical Modeling, and Observational Cosmology. I am particularly interested in combining tools and methodologies from the field of Machine Learning (automatic differentiation, deep generative models, score-based high-dimensional inference) with physical modeling for the analysis of Stage IV Cosmological Surveys.

I am currently co-Chair of the LSST Informatics & Statistics Science Collaboration (LSST ISSC) and actively involved in the LSST Dark Energy Science Collaboration (LSST DESC).

Before my CNRS position, I was a postdoctoral fellow at the Berkeley Center for Cosmological Physics (BCCP) and with the Foundation of Data Analysis (FODA) institute at UC Berkeley, working with Prof. Uroš Seljak. Before that, I was a postdoctoral researcher in the McWilliams Center for Cosmology at Carnegie Mellon University, where I was working with Prof. Rachel Mandelbaum on weak gravitational lensing measurements and systematics, and interacted with both Statistics and Machine Learning departments while at CMU.

news

Jan 30, 2023	Open PhD position at CEA Saclay to work with me on using generative modeling and autodiff to measure cosmic shear from space 🛰️. Application deadline March 10th.
Dec 12, 2022	Very cool project led by Yesukhei Jagvaral in collaboration with Rachel Mandelbaum on developing diffusion models for the SO3 Lie group to model galaxy intrinsic orientations in simulations. Full ML paper on this coming soon.
Nov 26, 2022	Pre-release of jaxDecomp, a JAX library providing bindings to NVIDIA’s cuDecomp adaptive pencil decomposition library for efficient multi-node/multi-gpu 3D FFTs and halo exchanges.
Nov 15, 2022	Multiple papers accepted in the Machine Learning and the Physical Sciences Workshop at NeurIPS 2022
Nov 7, 2022	New astro paper on Differentiable Halo Occupation Distributions led by Ben Horowitz demonstrating that it’s not because your model is stochastic and involves discrete variables that you can’t backprop through it
Feb 13, 2022	(jax-cosmo paper)[https://arxiv.org/abs/2302.05163] is out on the arxiv thanks to the work of a lot of people! Fisher forecasts are now officially super easy!

selected publications

Differentiable Stochastic Halo Occupation Distribution

Benjamin Horowitz, ChangHoon Hahn, Francois Lanusse, and 2 more authors

arXiv e-prints Nov 2022

Abs arXiv Bib Code

In this work, we demonstrate how differentiable stochastic sampling techniques developed in the context of deep Reinforcement Learning can be used to perform efficient parameter inference over stochastic, simulation-based, forward models. As a particular example, we focus on the problem of estimating parameters of Halo Occupancy Distribution (HOD) models which are used to connect galaxies with their dark matter halos. Using a combination of continuous relaxation and gradient parameterization techniques, we can obtain well-defined gradients with respect to HOD parameters through discrete galaxy catalogs realizations. Having access to these gradients allows us to leverage efficient sampling schemes, such as Hamiltonian Monte-Carlo, and greatly speed up parameter inference. We demonstrate our technique on a mock galaxy catalog generated from the Bolshoi simulation using the Zheng et al. 2007 HOD model and find near identical posteriors as standard Markov Chain Monte Carlo techniques with an increase of \raisebox-0.5ex\textasciitilde8x in convergence efficiency. Our differentiable HOD model also has broad applications in full forward model approaches to cosmic structure and cosmological analysis.
@article{2022arXiv221103852H, adsnote = {Provided by the SAO/NASA Astrophysics Data System}, adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv221103852H}, archiveprefix = {arXiv}, author = {{Horowitz}, Benjamin and {Hahn}, ChangHoon and {Lanusse}, Francois and {Modi}, Chirag and {Ferraro}, Simone}, eid = {arXiv:2211.03852}, eprint = {2211.03852}, journal = {arXiv e-prints}, keywords = {Astrophysics - Cosmology and Nongalactic Astrophysics, Astrophysics - Astrophysics of Galaxies}, month = nov, pages = {arXiv:2211.03852}, primaryclass = {astro-ph.CO}, title = {{Differentiable Stochastic Halo Occupation Distribution}}, year = {2022} }
The DAWES review 10: The impact of deep learning for the analysis of galaxy surveys

Marc Huertas-Company, and François Lanusse

arXiv e-prints Oct 2022

Abs arXiv Bib Website

The amount and complexity of data delivered by modern galaxy surveys has been steadily increasing over the past years. Extracting coherent scientific information from these large and multi-modal data sets remains an open issue and data driven approaches such as deep learning have rapidly emerged as a potentially powerful solution to some long lasting challenges. This enthusiasm is reflected in an unprecedented exponential growth of publications using neural networks. Half a decade after the first published work in astronomy mentioning deep learning, we believe it is timely to review what has been the real impact of this new technology in the field and its potential to solve key challenges raised by the size and complexity of the new datasets. In this review we first aim at summarizing the main applications of deep learning for galaxy surveys that have emerged so far. We then extract the major achievements and lessons learned and highlight key open questions and limitations. Overall, state-of-the art deep learning methods are rapidly adopted by the astronomical community, reflecting a democratization of these methods. We show that the majority of works using deep learning up to date are oriented to computer vision tasks. This is also the domain of application where deep learning has brought the most important breakthroughs so far. We report that the applications are becoming more diverse and deep learning is used for estimating galaxy properties, identifying outliers or constraining the cosmological model. Most of these works remain at the exploratory level. Some common challenges will most likely need to be addressed before moving to the next phase of deployment of deep learning in the processing of future surveys; e.g. uncertainty quantification, interpretability, data labeling and domain shift issues from training with simulations, which constitutes a common practice in astronomy.
@article{2022arXiv221001813H, adsnote = {Provided by the SAO/NASA Astrophysics Data System}, adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv221001813H}, archiveprefix = {arXiv}, author = {{Huertas-Company}, Marc and {Lanusse}, Fran{\c{c}}ois}, eid = {arXiv:2210.01813}, eprint = {2210.01813}, journal = {arXiv e-prints}, keywords = {Astrophysics - Instrumentation and Methods for Astrophysics, Astrophysics - Cosmology and Nongalactic Astrophysics, Astrophysics - Astrophysics of Galaxies}, month = oct, pages = {arXiv:2210.01813}, primaryclass = {astro-ph.IM}, title = {{The DAWES review 10: The impact of deep learning for the analysis of galaxy surveys}}, year = {2022} }
Nat. Rev. Phys
Bayesian uncertainty quantification for machine-learned models in physics

Yarin Gal, Petros Koumoutsakos, Francois Lanusse, and 2 more authors

Nature Reviews Physics Aug 2022

Abs Bib

Being able to quantify uncertainty when comparing a theoretical or computational model to observations is critical to conducting a sound scientific investigation. With the rise of data-driven modelling, understanding various sources of uncertainty and developing methods to estimate them has gained renewed attention. Five researchers discuss uncertainty quantification in machine-learned models with an emphasis on issues relevant to physics problems.
@article{2022NatRP...4..573G, adsnote = {Provided by the SAO/NASA Astrophysics Data System}, adsurl = {https://ui.adsabs.harvard.edu/abs/2022NatRP...4..573G}, author = {{Gal}, Yarin and {Koumoutsakos}, Petros and {Lanusse}, Francois and {Louppe}, Gilles and {Papadimitriou}, Costas}, doi = {10.1038/s42254-022-00498-4}, journal = {Nature Reviews Physics}, month = aug, number = {9}, pages = {573-577}, title = {{Bayesian uncertainty quantification for machine-learned models in physics}}, volume = {4}, year = {2022} }
Neural Posterior Estimation with Differentiable Simulators

Justine Zeghal, François Lanusse, Alexandre Boucaud, and 2 more authors

arXiv e-prints Jul 2022

Abs arXiv Bib Slides

Simulation-Based Inference (SBI) is a promising Bayesian inference framework that alleviates the need for analytic likelihoods to estimate posterior distributions. Recent advances using neural density estimators in SBI algorithms have demonstrated the ability to achieve high-fidelity posteriors, at the expense of a large number of simulations ; which makes their application potentially very time-consuming when using complex physical simulations. In this work we focus on boosting the sample- efficiency of posterior density estimation using the gradients of the simulator. We present a new method to perform Neural Posterior Estimation (NPE) with a differentiable simulator. We demonstrate how gradient information helps constrain the shape of the posterior and improves sample-efficiency.
@article{2022arXiv220705636Z, adsnote = {Provided by the SAO/NASA Astrophysics Data System}, adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv220705636Z}, archiveprefix = {arXiv}, author = {{Zeghal}, Justine and {Lanusse}, Fran{\c{c}}ois and {Boucaud}, Alexandre and {Remy}, Benjamin and {Aubourg}, Eric}, eid = {arXiv:2207.05636}, eprint = {2207.05636}, journal = {arXiv e-prints}, keywords = {Astrophysics - Instrumentation and Methods for Astrophysics, Statistics - Machine Learning}, month = jul, pages = {arXiv:2207.05636}, primaryclass = {astro-ph.IM}, title = {{Neural Posterior Estimation with Differentiable Simulators}}, year = {2022} }
Probabilistic Mass Mapping with Neural Score Estimation

Benjamin Remy, Francois Lanusse, Niall Jeffrey, and 4 more authors

arXiv e-prints Jan 2022

Abs arXiv Bib Code

Weak lensing mass-mapping is a useful tool to access the full distribution of dark matter on the sky, but because of intrinsic galaxy ellipticies and finite fields/missing data, the recovery of dark matter maps constitutes a challenging ill-posed inverse problem. We introduce a novel methodology allowing for efficient sampling of the high-dimensional Bayesian posterior of the weak lensing mass-mapping problem, and relying on simulations for defining a fully non-Gaussian prior. We aim to demonstrate the accuracy of the method on simulations, and then proceed to applying it to the mass reconstruction of the HST/ACS COSMOS field. The proposed methodology combines elements of Bayesian statistics, analytic theory, and a recent class of Deep Generative Models based on Neural Score Matching. This approach allows us to do the following: 1) Make full use of analytic cosmological theory to constrain the 2pt statistics of the solution. 2) Learn from cosmological simulations any differences between this analytic prior and full simulations. 3) Obtain samples from the full Bayesian posterior of the problem for robust Uncertainty Quantification. We demonstrate the method on the \kappaTNG simulations and find that the posterior mean significantly outperfoms previous methods (Kaiser-Squires, Wiener filter, Sparsity priors) both on root-mean-square error and in terms of the Pearson correlation. We further illustrate the interpretability of the recovered posterior by establishing a close correlation between posterior convergence values and SNR of clusters artificially introduced into a field. Finally, we apply the method to the reconstruction of the HST/ACS COSMOS field and yield the highest quality convergence map of this field to date.
@article{2022arXiv220105561R, adsnote = {Provided by the SAO/NASA Astrophysics Data System}, adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv220105561R}, archiveprefix = {arXiv}, author = {{Remy}, Benjamin and {Lanusse}, Francois and {Jeffrey}, Niall and {Liu}, Jia and {Starck}, Jean-Luc and {Osato}, Ken and {Schrabback}, Tim}, eid = {arXiv:2201.05561}, eprint = {2201.05561}, journal = {arXiv e-prints}, keywords = {Astrophysics - Cosmology and Nongalactic Astrophysics, Astrophysics - Instrumentation and Methods for Astrophysics, Computer Science - Machine Learning}, month = jan, pages = {arXiv:2201.05561}, primaryclass = {astro-ph.CO}, title = {{Probabilistic Mass Mapping with Neural Score Estimation}}, year = {2022} }