arXiv subjects

Pierre Chanial

Publications and source records attributed to Pierre Chanial.

11 recordsLinked to original sources

arXiv2026

A GPU-Accelerated JAX Framework for Robust Parametric Component Separation and Clustering Optimization for CMB Polarization Satellites

We present a novel, JAX-powered implementation of a parametric component-separation method for CMB polarization data, explicitly designed to handle spatially varying foreground Spectral Energy Distributions (SEDs). The approach models this variation across the sky by grouping sets of pixels that share common foreground spectral parameters, scanning over thousands of such configurations to evaluate the trade-off between model complexity and residual systematic contamination. Built within the FURAX framework -- a JAX-powered environment for CMB data analysis -- our pipeline extends the fgbuster parametric formalism. It enables fully vectorized, GPU-accelerated evaluation of the spectral likelihood, map reconstruction, and diagnostic metrics across tens of thousands of pixel subset configurations, noise realizations, and sky regions. Our implementation achieves up to $\sim 100\times$ speed-up over the scipy TNC optimizer used in fgbuster when running on GPUs, as well as giving more robust results. When applied to LiteBIRD-like simulations with spatially varying foreground SEDs, our optimized K-means configuration reduces the 68% upper limit on the tensor-to-scalar ratio $r$ by $\approx 30\%$ relative to a fixed, previously derived multi-resolution configuration, while maintaining competitive statistical uncertainties.

astro-ph.CO↗

arXiv2026

Furax: A Modular JAX Framework for Linear Operators in Astrophysical and Cosmological Data Analysis

The Framework for Unified and Robust data Analysis with JAX (Furax) is an open-source Python framework for modeling data acquisition systems and solving inverse problems in astrophysics and cosmology. Built on JAX, Furax provides composable building blocks in the form of general-purpose and domain-specific linear operators, along with preconditioners and solvers for their numerical inversion. Domain-specific tools are provided for astrophysical and cosmic microwave background (CMB) data analysis$-$including map-making, instrument modeling, and astrophysical component separation$-$with a modular architecture designed to extend to other fields.

astro-ph.IM↗

arXiv2026

Half-wave-plate non idealities propagated to component separated CMB $B$-modes

We assess the impact of non-ideal, continuously rotating half-wave plates (HWPs) on cosmic microwave background (CMB) polarization measurements targeting large angular scale signal. Such hardware solutions are used in or planned for multiple modern CMB efforts, both ground-based, for instance, small aperture telescopes of Simons Observatory or satellite borne, such as LiteBIRD. Using a frequency-dependent parametric model based on the Mueller matrix formalism, we characterize the induced mixing of Stokes parameters. Through end-to-end simulations, we propagate these effects from time-ordered data to cosmology via map-making and component-separation stages, quantifying their impact on the $B$-modes power spectrum and the tensor-to-scalar ratio, $r$. Our analysis shows that neglecting the frequency dependence of a three-layer HWP gives rise to significant polarization leakage, biases foreground spectral parameters, and leads to residual contamination in the recovered CMB maps. To mitigate these effects, we investigate multiple analysis strategies progressively incorporating a more complete description of the instrumental response. At the map-making level, this requires generalizing the standard pointing matrix to account for the full time- and frequency-dependent instrumental response. We find that standard HWP models, reduce the biases only down to $r \sim 10^{-2}$, while a more advanced approach based on a generalization of both map-making and component separation, implemented using JAX, can suppress it down to $r \sim 7 \times 10^{-4}$. Finally, we extend this approach to a time-domain component-separation, enabling a statistically consistent treatment of instrumental response in the presence of time-domain features. We demonstrate its feasibility and validate it by performing a full end-to-end analysis, recovering results in good agreement with the map-based ones.

astro-ph.CO↗

arXiv2021

Markov Decision Process for MOOC users behavioral inference

Studies on massive open online courses (MOOCs) users discuss the existence of typical profiles and their impact on the learning process of the students. However defining the typical behaviors as well as classifying the users accordingly is a difficult task. In this paper we suggest two methods to model MOOC users behaviour given their log data. We mold their behavior into a Markov Decision Process framework. We associate the user's intentions with the MDP reward and argue that this allows us to classify them.

cs.LG↗

arXiv2019

QUBIC: Exploring the primordial Universe with the Q\&U Bolometric Interferometer

In this paper we describe QUBIC, an experiment that will observe the polarized microwave sky with a novel approach, which combines the sensitivity of state-of-the art bolometric detectors with the systematic effects control typical of interferometers. QUBIC unique features are the so-called "self-calibration", a technique that allows us to clean the measured data from instrumental effects, and its spectral imaging power, i.e. the ability to separate the signal in various sub-bands within each frequency band. QUBIC will observe the sky in two main frequency bands: 150 GHz and 220 GHz. A technological demonstrator is currently under testing and will be deployed in Argentina during 2019, while the final instrument is expected to be installed during 2020.

astro-ph.IM↗

arXiv2016

A Deconvolution Technique to Correct Deep Images of Galaxies from Instrumental Scattered Light

Deep imaging of the diffuse light emitted by the stellar fine structures and outer halos around galaxies is now often used to probe their past mass assembly. Because the extended halos survive longer than the relatively fragile tidal features, they trace more ancient mergers. We use images reaching surface brightness limits as low as 28.5-29 mag.arcsec-2 (g-band) to obtain light and color profiles up to 5-10 effective radii of a sample of nearby early-type galaxies. They were acquired with MegaCam as part of the CFHT MATLAS large programme. These profiles may be compared to those produced by simulations of galaxy formation and evolution, once corrected for instrumental effects. Indeed they can be heavily contaminated by the scattered light caused by internal reflections within the instrument. In particular, the nucleus of galaxies generates artificial flux in the outer halo, which has to be precisely subtracted. We present a deconvolution technique to remove the artificial halos that makes use of very large kernels. The technique based on PyOperators is more time efficient than the model-convolution methods also used for that purpose. This is especially the case for galaxies with complex structures that are hard to model. Having a good knowledge of the Point Spread Function (PSF), including its outer wings, is critical for the method. A database of MegaCam PSF models corresponding to different seeing conditions and bands was generated directly from the deep images. It is shown that the difference in the PSFs in different bands causes artificial changes in the color profiles, in particular a reddening of the outskirts of galaxies having a bright nucleus. The method is validated with a set of simulated images and applied to three representative test cases: NGC 3599, NGC 3489, and NGC 4274, and exhibiting for two of them a prominent ghost halo. The method successfully removes it.

astro-ph.GA↗

arXiv2010

Feasibility and performances of compressed-sensing and sparse map-making with Herschel/PACS data

The Herschel Space Observatory of ESA was launched in May 2009 and is in operation since. From its distant orbit around L2 it needs to transmit a huge quantity of information through a very limited bandwidth. This is especially true for the PACS imaging camera which needs to compress its data far more than what can be achieved with lossless compression. This is currently solved by including lossy averaging and rounding steps on board. Recently, a new theory called compressed-sensing emerged from the statistics community. This theory makes use of the sparsity of natural (or astrophysical) images to optimize the acquisition scheme of the data needed to estimate those images. Thus, it can lead to high compression factors. A previous article by Bobin et al. (2008) showed how the new theory could be applied to simulated Herschel/PACS data to solve the compression requirement of the instrument. In this article, we show that compressed-sensing theory can indeed be successfully applied to actual Herschel/PACS data and give significant improvements over the standard pipeline. In order to fully use the redundancy present in the data, we perform full sky map estimation and decompression at the same time, which cannot be done in most other compression methods. We also demonstrate that the various artifacts affecting the data (pink noise, glitches, whose behavior is a priori not well compatible with compressed-sensing) can be handled as well in this new framework. Finally, we make a comparison between the methods from the compressed-sensing scheme and data acquired with the standard compression scheme. We discuss improvements that can be made on ground for the creation of sky maps from the data.

astro-ph.IM↗

arXiv2007

First Constraints on Source Counts at 350 Microns

We have imaged a $\sim$6 arcminute$^2$ region in the Boötes Deep Field using the 350 $μ$m-optimised second generation Submillimeter High Angular Resolution Camera (SHARC II), achieving a peak 1$σ$ sensitivity of $\sim$5 mJy. We detect three sources above 3$σ$, and determine a spurious source detection rate of 1.09 in our maps. In the absence of $5σ$ detections, we rely on deep 24 $μ$m and 20 cm imaging to deduce which sources are most likely to be genuine, giving two real sources. From this we derive an integral source count of 0.84$^{+1.39}_{-0.61}$ sources arcmin$^{-2}$ at $S>13$ mJy, which is consistent with 350 $μ$m source count models that have an IR-luminous galaxy population evolving with redshift. We use these constraints to consider the future for ground-based short-submillimetre surveys.

astro-ph↗

arXiv2006

The infrared compactness-temperature relation for quiescent and starburst galaxies

IRAS observations show the existence of a correlation between the infrared luminosity Lir and dust temperature Td in star-forming galaxies, in which larger Lir leads to higher dust temperature. The Lir-Td relation is commonly seen as reflecting the increase in dust temperature in galaxies with higher star formation rate. Even though the correlation shows a significant amount of dispersion, a unique relation has been commonly used to construct spectral energy distributions of galaxies in distant universe studies, such as source number counting or photometric redshift determination. In this work, we introduce a new parameter, namely the size of the star-forming region Rir and lay out the empirical and modelled relation between the global parameters Lir, Td and Rir of IR-bright non-AGN galaxies. IRAS 60-to-100um color is used as a proxy for the dust temperature and the 1.4GHz radio contiuum emission for the infrared spatial distribution. The analysis has been carried out on two samples. The first one is made of the galaxies from the 60um flux-limited IRAS Revised Bright Galaxy Samples which have a reliable RC size estimate from the VLA follow-ups of the IRAS Bright Galaxy Samples. The second is made of the sources from the 170um ISOPHOT Serendipity Sky Survey which are resolved by the NVSS or FIRST surveys. We show that the dispersion in the Lir-Td diagram can be reduced to a relation between the infrared surface brightness and the dust temperature, a relation that spans 5 orders of magnitude in surface brightness. We explored the physical processes giving rise to the Sir-Td relation, and show that it can be derived from the Schmidt law, which relates the star formation rate to the gas surface density.

astro-ph↗

arXiv2005

On the discovery of the first 350 micron-selected galaxy

We report the detection of a 3.6sigma 350micron-selected source in the Bootes Deep Field. The source, the first Short-wavelength Submillimeter-selected Galaxy (SSG 1), was discovered as part of a blank field extragalactic survey using the 350micron-optimised Submillimeter High Angular Resolution Camera (SHARC II) at the Caltech Submillimeter Observatory. With multiwavelength photometry from NOAO-NDWFS (R and I band), FLAMEX (J and K_s), Spitzer (IRAC and MIPS) and the Westerbork 1.4GHz Deep Survey (radio upper limit), we are able to constrain the photometric redshift using different methods, all of which suggest a redshift of close to 1. In the absence of long-wavelength submillimeter data we use SED templates to infer that this source is an ultraluminous infrared galaxy (ULIRG) with a dust temperature of 30+/-5 K, occupying a region of luminosity-temperature space shared by modarate redshift ISO-selected ULIRGs (rather than high redshift SCUBA-selected SMGs). SHARC II can thus select SMGs with moderately "warm'' dust that might be missed in submillimeter surveys at longer wavelengths.

astro-ph↗

arXiv2001

ISOCAM Mid-InfraRed Detection of HR 10: A Distant Clone of Arp 220 at z=1.44

We report the detection of the extremely red object (ERO), HR 10 (I-K= 6.5, z=1.44), at 4.9 and 6.1 microns (rest-frame) with ISOCAM, the mid-infrared (MIR) camera onboard the Infrared Space Observatory (ISO). HR 10 is the first ERO spectroscopically identified to be associated with an ultra-luminous IR galaxy (ULIG) detected in the radio, MIR and sub-millimeter. The rest-frame spectral energy distribution (SED) of HR 10 is amazingly similar to the one of Arp 220, scaled by a factor 3.8+/-1.3. The corresponding 8-1000 microns luminosity (~ 7x10^12 h70^{-2} Lsol) translates into a star formation rate of about 1200 h70^{-2} Msol/yr if HR 10 is mostly powered by star formation. We address the key issue of the origin of the powerful luminosity of HR 10, i.e. starburst versus active galactic nucleus (AGN), by using the similarity with its closeby clone, Arp 220.

astro-ph↗