arXiv subjects

Brice Ménard

Publications and source records attributed to Brice Ménard.

At least 19 recordsLinked to original sources

arXiv2026

Tomographer: End-to-end Redshift Distribution Estimation for Source Catalogs and Intensity Maps

Redshift information is central to nearly every extragalactic and cosmological application of sky surveys, yet only a small fraction of cataloged sources, and none of the photons forming diffuse backgrounds, have spectroscopic redshifts. Clustering-based redshift inference estimates the redshift distribution of an arbitrary dataset through spatial cross-correlation with a reference sample of known redshifts. It relies only on positional information, making it applicable to any tracer of large-scale structure, including source populations and diffuse intensity maps. Its broader adoption, however, has been limited by technical barriers: assembling and characterizing spectroscopic references, expensive pair-counting computations, theoretical corrections, and systematic control. To remove these barriers, we introduce Tomographer, an end-to-end clustering-redshift framework. The key design feature is a set of precomputed "activation maps" encoding the spatial pair information of 3 million SDSS spectroscopic galaxies and quasars up to $z\sim4$ over $10{,}000\,{\rm deg}^2$. This eliminates user-end pair counting, reducing computational scaling from $O(N\log N)$ to $O(1)$ map multiplications. Given a source catalog or intensity map, Tomographer returns the bias-weighted redshift distribution, $b(z)\,{\rm d}N/{\rm d}z(z)$ or $b(z)\,{\rm d}I/{\rm d}z(z)$. We validate the framework against samples with known redshifts, demonstrate accurate uncertainty estimates, and show robustness to survey footprint, spatially varying selection functions, beam smoothing, and foreground contamination. We showcase applications to source catalogs selected by flux, color, photometric redshift, morphology, or variability, and to intensity maps from radio to X-rays. Future Tomographer releases will incorporate additional wide-field spectroscopic reference samples as they become available.

astro-ph.CO↗

arXiv2026

The Future of Artificial Intelligence and the Mathematical and Physical Sciences (AI+MPS)

This community paper developed out of the NSF Workshop on the Future of Artificial Intelligence (AI) and the Mathematical and Physics Sciences (MPS), which was held in March 2025 with the goal of understanding how the MPS domains (Astronomy, Chemistry, Materials Research, Mathematical Sciences, and Physics) can best capitalize on, and contribute to, the future of AI. We present here a summary and snapshot of the MPS community's perspective, as of Spring/Summer 2025, in a rapidly developing field. The link between AI and MPS is becoming increasingly inextricable; now is a crucial moment to strengthen the link between AI and Science by pursuing a strategy that proactively and thoughtfully leverages the potential of AI for scientific discovery and optimizes opportunities to impact the development of AI by applying concepts from fundamental science. To achieve this, we propose activities and strategic priorities that: (1) enable AI+MPS research in both directions; (2) build up an interdisciplinary community of AI+MPS researchers; and (3) foster education and workforce development in AI for MPS researchers and students. We conclude with a summary of suggested priorities for funding agencies, educational institutions, and individual researchers to help position the MPS community to be a leader in, and take full advantage of, the transformative potential of AI+MPS.

cs.AI↗

arXiv2025

Universal scale-free representations in human visual cortex

How does the human brain encode complex visual information? While previous research has characterized individual dimensions of visual representation in cortex, we still lack a comprehensive understanding of how visual information is organized across the full range of neural population activity. Here, analyzing fMRI responses to natural scenes across multiple individuals, we discover that neural representations in human visual cortex follow a remarkably consistent scale-free organization -- their variance systematically decays as a power law, detected across four orders of magnitude of latent dimensions. This scale-free structure appears consistently across multiple visual regions and across individuals, suggesting it reflects a fundamental organizing principle of visual processing. Critically, when we align neural responses across individuals using hyperalignment, we find that these representational dimensions are largely shared between people, revealing a universal high-dimensional spectrum of visual information that emerges despite individual differences in brain anatomy and visual experience. Traditional analysis approaches in cognitive neuroscience have focused primarily on a small number of high-variance dimensions, potentially missing crucial aspects of visual representation. Our results demonstrate that visual information is distributed across the full dimensionality of cortical activity in a systematic way, suggesting we need to move beyond low-dimensional characterizations to fully understand how the brain represents the visual world. This work reveals a new fundamental principle of neural coding in human visual cortex and highlights the importance of examining neural representations across their full dimensionality.

q-bio.NC↗

arXiv2025

Cosmic Infrared Background Tomography and a Census of Cosmic Dust and Star Formation

The cosmic far-infrared background (CIB) encodes dust emission from all galaxies and carries valuable information on structure formation, star formation, and chemical enrichment across cosmic time. However, its redshift-dependent spectrum remains poorly constrained due to line-of-sight projection effects. We address this by cross-correlating 11 far-infrared intensity maps spanning a 50-fold frequency range from Planck, Herschel, and IRAS, with spectroscopic galaxies and quasars from SDSS I-IV tomographically. We mitigate foregrounds using CSFD, a CIB-free Milky Way dust map. These cross-correlation amplitudes on two-halo scales trace bias-weighted CIB redshift distributions and collectively yield a $60σ$ detection of the evolving CIB spectrum, sampled across hundreds of rest-frame frequencies over $0 < z < 4$. We break the bias-intensity degeneracy by adding monopole information from FIRAS+Planck. The recovered spectrum reveals a dust temperature distribution that is broad, spanning the full range of host environments, and moderately evolving. Using low-frequency CIB amplitudes, we constrain cosmic dust density, $Ω_{\rm dust}$, which peaks at $z = 1$-$1.5$ and declines threefold to the present. Our broad spectral coverage enables a determination of the total infrared luminosity density to 0.04 dex statistical precision, tracing star-formation history with negligible cosmic variance across 90% of cosmic time. We find that cosmic star formation is 80% dust-obscured at $z = 0$ and 60% at $z = 4$. Our results, based on intensity mapping, are complete, requiring no extrapolation to faint galaxies or low-surface-brightness components. We release our tomographic CIB spectrum and redshift distributions as a public resource for future studies of the CIB, both as a cosmological matter tracer and CMB foreground.

astro-ph.CO↗

arXiv2024

A Rainbow in Deep Network Black Boxes

A central question in deep learning is to understand the functions learned by deep networks. What is their approximation class? Do the learned weights and representations depend on initialization? Previous empirical work has evidenced that kernels defined by network activations are similar across initializations. For shallow networks, this has been theoretically studied with random feature models, but an extension to deep networks has remained elusive. Here, we provide a deep extension of such random feature models, which we call the rainbow model. We prove that rainbow networks define deterministic (hierarchical) kernels in the infinite-width limit. The resulting functions thus belong to a data-dependent RKHS which does not depend on the weight randomness. We also verify numerically our modeling assumptions on deep CNNs trained on image classification tasks, and show that the trained networks approximately satisfy the rainbow hypothesis. In particular, rainbow networks sampled from the corresponding random feature model achieve similar performance as the trained networks. Our results highlight the central role played by the covariances of network weights at each layer, which are observed to be low-rank as a result of feature learning.

cs.LG↗

arXiv2024

Scattering Spectra Models for Physics

Physicists routinely need probabilistic models for a number of tasks such as parameter inference or the generation of new realizations of a field. Establishing such models for highly non-Gaussian fields is a challenge, especially when the number of samples is limited. In this paper, we introduce scattering spectra models for stationary fields and we show that they provide accurate and robust statistical descriptions of a wide range of fields encountered in physics. These models are based on covariances of scattering coefficients, i.e. wavelet decomposition of a field coupled with a point-wise modulus. After introducing useful dimension reductions taking advantage of the regularity of a field under rotation and scaling, we validate these models on various multi-scale physical fields and demonstrate that they reproduce standard statistics, including spatial moments up to 4th order. These scattering spectra provide us with a low-dimensional structured representation that captures key properties encountered in a wide range of physical fields. These generic models can be used for data exploration, classification, parameter inference, symmetry detection, and component separation.

physics.data-an↗

arXiv2024

On the universality of neural encodings in CNNs

We explore the universality of neural encodings in convolutional neural networks trained on image classification tasks. We develop a procedure to directly compare the learned weights rather than their representations. It is based on a factorization of spatial and channel dimensions and measures the similarity of aligned weight covariances. We show that, for a range of layers of VGG-type networks, the learned eigenvectors appear to be universal across different natural image datasets. Our results suggest the existence of a universal neural encoding for natural images. They explain, at a more fundamental level, the success of transfer learning. Our work shows that, instead of aiming at maximizing the performance of neural networks, one can alternatively attempt to maximize the universality of the learned encoding, in order to build a principled foundation model.

cs.LG↗

arXiv2024

Cosmological constraints from weak lensing scattering transform using HSC Y1 data

As weak lensing surveys go deeper, there is an increasing need for reliable characterization of non-Gaussian structures at small angular scales. Here we present the first cosmological constraints with weak lensing scattering transform, a statistical estimator that combines efficiency, robustness, and interpretability. With the Hyper Suprime-Cam survey (HSC) year 1 data, we obtain $Ω_\text{m}=0.29_{-0.03}^{+0.04}$, $S_8\equiv σ_8(Ω_\text{m}/0.3)^{0.5}=0.83\pm0.02$, and intrinsic alignment strength $A_\text{IA}=1.0\pm0.4$ through simulation-based forward modeling. Our constraints are consistent with those derived from Planck. The error bar of $Ω_\text{m}$ is 2 times tighter than that obtained from the power spectrum when the same scale range is used. This constraining power is on par with that of convolutional neural networks, suggesting that further investment in spatial information extraction may not yield substantial benefits. We also point out an internal tension of $S_8$ estimates linked to a redshift bin around z ~ 1 in the HSC data. We found that discarding that bin leads to a consistent decrease of $S_8$ from 0.83 to 0.79, for all statistical estimators. We argue that photometric redshift estimation is now the main limitation in the estimation of $S_8$ using HSC. This limitation is likely to affect other ground-based weak lensing surveys reaching redshifts greater than one. Alternative redshift estimation techniques, like clustering redshifts, may help alleviate this limitation.

astro-ph.CO↗

arXiv2021

How to quantify fields or textures? A guide to the scattering transform

Extracting information from stochastic fields or textures is a ubiquitous task in science, from exploratory data analysis to classification and parameter estimation. From physics to biology, it tends to be done either through a power spectrum analysis, which is often too limited, or the use of convolutional neural networks (CNNs), which require large training sets and lack interpretability. In this paper, we advocate for the use of the scattering transform (Mallat 2012), a powerful statistic which borrows mathematical ideas from CNNs but does not require any training, and is interpretable. We show that it provides a relatively compact set of summary statistics with visual interpretation and which carries most of the relevant information in a wide range of scientific applications. We present a non-technical introduction to this estimator and we argue that it can benefit data analysis, comparison to models and parameter inference in many fields of science. Interestingly, understanding the core operations of the scattering transform allows one to decipher many key aspects of the inner workings of CNNs.

astro-ph.IM↗

arXiv2021

Weak lensing scattering transform: dark energy and neutrino mass sensitivity

As weak lensing surveys become deeper, they reveal more non-Gaussian aspects of the convergence field which can only be extracted using statistics beyond the power spectrum. In Cheng et al. (2020) we showed that the scattering transform, a novel statistic borrowing mathematical concepts from convolutional neural networks, is a powerful tool for cosmological parameter estimation in the non-Gaussian regime. Here, we extend that analysis to explore its sensitivity to dark energy and neutrino mass parameters with weak lensing surveys. We first use image synthesis to show visually that, compared to the power spectrum and bispectrum, the scattering transform provides a better statistical vocabulary to characterize the perceptual properties of lensing mass maps. We then show that it is also better suited for parameter inference: (i) it provides higher sensitivity in the noiseless regime, and (ii) at the noise level of Rubin-like surveys, though the constraints are not significantly tighter than those of the bispectrum, the scattering coefficients have a more Gaussian sampling distribution, which is an important property for likelihood parametrization and accurate cosmological inference. We argue that the scattering coefficients are preferred statistics considering both constraining power and likelihood properties.

astro-ph.CO↗

arXiv2021

The thermal and gravitational energy densities in the large-scale structure of the Universe

As cosmic structures form, matter density fluctuations collapse gravitationally and baryonic matter is shock-heated and thermalized. We therefore expect a connection between the mean gravitational potential energy density of collapsed halos, $Ω_{W}^{\rm halo}$, and the mean thermal energy density of baryons, $Ω_{\rm th}$. These quantities can be obtained using two fundamentally different estimates: we compute $Ω_{W}^{\rm halo}$ using the theoretical framework of the halo model which is driven by dark matter statistics, and measure $Ω_{\rm th}$ using the Sunyaev-Zeldovich (SZ) effect which probes the mean thermal pressure of baryons. First, we derive that, at the present time, about 90% of $Ω_{W}^{\rm halo}$ originates from massive halos with $M>10^{13}\,M_\odot$. Then, using our measurements of the SZ background, we find that $Ω_{\rm th}$ accounts for about 80% of the kinetic energy of the baryons available for pressure in halos at $z\lesssim 0.5$. This constrains the amount of non-thermal pressure, e.g., due to bulk and turbulent gas motion sourced by mass accretion, to be about $Ω_{\rm non-th}\simeq 0.4\times 10^{-8}$ at $z=0$.

astro-ph.CO↗

arXiv2020

A new approach to observational cosmology using the scattering transform

Parameter estimation with non-Gaussian stochastic fields is a common challenge in astrophysics and cosmology. In this paper, we advocate performing this task using the scattering transform, a statistical tool sharing ideas with convolutional neural networks (CNNs) but requiring no training nor tuning. It generates a compact set of coefficients, which can be used as robust summary statistics for non-Gaussian information. It is especially suited for fields presenting localized structures and hierarchical clustering, such as the cosmological density field. To demonstrate its power, we apply this estimator to a cosmological parameter inference problem in the context of weak lensing. On simulated convergence maps with realistic noise, the scattering transform outperforms classic estimators and is on a par with state-of-the-art CNN. It retains the advantages of traditional statistical descriptors, has provable stability properties, allows to check for systematics, and importantly, the scattering coefficients are interpretable. It is a powerful and attractive estimator for observational cosmology and the study of physical fields in general.

astro-ph.CO↗

arXiv2020

The Cosmic Thermal History Probed by Sunyaev-Zeldovich Effect Tomography

The cosmic thermal history, quantified by the evolution of the mean thermal energy density in the universe, is driven by the growth of structures as baryons get shock heated in collapsing dark matter halos. This process can be probed by redshift-dependent amplitudes of the thermal Sunyaev-Zeldovich (SZ) effect background. To do so, we cross-correlate eight sky intensity maps in the $\it{Planck}$ and Infrared Astronomical Satellite missions with two million spectroscopic redshift references in the Sloan Digital Sky Surveys. This delivers snapshot spectra for the far-infrared to microwave background light as a function of redshift up to $z\sim3$. We decompose them into the SZ and thermal dust components. Our SZ measurements directly constrain $\langle bP_{\rm e} \rangle$, the halo bias-weighted mean electron pressure, up to $z\sim 1$. This is the highest redshift achieved to date, with uncorrelated redshift bins thanks to the spectroscopic references. We detect a threefold increase in the density-weighted mean electron temperature $\bar{T}_{\rm{e}}$ from $7\times 10^5~{\rm K}$ at $z=1$ to $2\times 10^6~{\rm K}$ today. Over $z=1$-$0$, we witness the build-up of nearly $70\%$ of the present-day mean thermal energy density $ρ_{\rm{th}}$, with the corresponding density parameter $Ω_{\rm th}$ reaching $1.5 \times10^{-8}$. We find the mass bias parameter of $\it{Planck}$'s universal pressure profile of $B=1.27$ (or $1-b=1/B=0.79$), consistent with the magnitude of non-thermal pressure in gas motion and turbulence from mass assembly. We estimate the redshift-integrated mean Compton parameter $y\sim1.2\times10^{-6}$, which will be tested by future spectral distortion experiments. More than half of which originates from the large-scale structure at $z<1$, which we detect directly.

astro-ph.CO↗

arXiv2020

Sequencing seismograms: A panoptic view of scattering in the core-mantle boundary region

Scattering of seismic waves can reveal subsurface structures but usually in a piecemeal way focused on specific target areas. We used a manifold learning algorithm called "the Sequencer" to simultaneously analyze thousands of seismograms of waves diffracting along the core-mantle boundary and obtain a panoptic view of scattering across the Pacific region. In nearly half of the diffracting waveforms, we detected seismic waves scattered by three-dimensional structures near the core-mantle boundary. The prevalence of these scattered arrivals shows that the region hosts pervasive lateral heterogeneity. Our analysis revealed loud signals due to a plume root beneath Hawaii and a previously unrecognized ultralow-velocity zone beneath the Marquesas Islands. These observations illustrate how approaches flexible enough to detect robust patterns with little to no user supervision can reveal distinctive insights into the deep Earth.

physics.geo-ph↗

arXiv2020

Extracting the main trend in a dataset: the Sequencer algorithm

Scientists aim to extract simplicity from observations of the complex world. An important component of this process is the exploration of data in search of trends. In practice, however, this tends to be more of an art than a science. Among all trends existing in the natural world, one-dimensional trends, often called sequences, are of particular interest as they provide insights into simple phenomena. However, some are challenging to detect as they may be expressed in complex manners. We present the Sequencer, an algorithm designed to generically identify the main trend in a dataset. It does so by constructing graphs describing the similarities between pairs of observations, computed with a set of metrics and scales. Using the fact that continuous trends lead to more elongated graphs, the algorithm can identify which aspects of the data are relevant in establishing a global sequence. Such an approach can be used beyond the proposed algorithm and can optimize the parameters of any dimensionality reduction technique. We demonstrate the power of the Sequencer using real-world data from astronomy, geology as well as images from the natural world. We show that, in a number of cases, it outperforms the popular t-SNE and UMAP dimensionality reduction techniques. This approach to exploratory data analysis, which does not rely on training nor tuning of any parameter, has the potential to enable discoveries in a wide range of scientific domains. The source code is available on github and we provide an online interface at \url{http://sequencer.org}.

cs.LG↗

arXiv2020

Double white dwarf merger products among high-mass white dwarfs

Double white dwarf (double-WD) binaries may merge within a Hubble time and produce high-mass WDs. Compared to other high-mass WDs, the double-WD merger products have higher velocity dispersion because they are older. With the power of Gaia data, we show strong evidence for double-WD merger products among high-mass WDs by analyzing the transverse-velocity distribution of more than a thousand high-mass WDs (0.8--1.3 $M_\odot$). We estimate that the fraction of double-WD merger products in our sample is about 20 %. We also obtain a precise double-WD merger rate and its mass dependence. Our merger rate estimates are close to binary population synthesis results and support the idea that double-WD mergers may contribute to a significant fraction of type Ia supernovae.

astro-ph.SR↗

arXiv2019

A cooling anomaly of high-mass white dwarfs

Recently, the power of Gaia data has revealed an enhancement of high-mass white dwarfs (WDs) on the Hertzsprung--Russell diagram, called the Q branch. This branch is located at the high-mass end of the recently identified crystallization branch. Investigating its properties, we find that the number density and velocity distribution on the Q branch cannot be explained by the cooling delay of crystallization alone, suggesting the existence of an extra cooling delay. To quantify this delay, we statistically compare two age indicators -- the dynamical age inferred from transverse velocity, and the photometric isochrone age -- for more than one thousand high-mass WDs (1.08--1.23 $M_\odot$) selected from Gaia Data Release 2. We show that about 6 % of the high-mass WDs must experience an 8 Gyr extra cooling delay on the Q branch, in addition to the crystallization and merger delays. This cooling anomaly is a challenge for WD cooling models. We point out that $^{22}$Ne settling in C/O-core WDs could account for this extra cooling delay.

astro-ph.SR↗

arXiv2019

Broadband Intensity Tomography: Spectral Tagging of the Cosmic UV Background

Cosmic photons can be efficiently collected by broadband intensity mapping but information on their emission redshift and frequency is largely lost. We introduce a technique to statistically recover these otherwise collapsed dimensions by exploiting information in spatial fluctuations and apply it to the Galaxy Evolution Explorer (GALEX) All Sky and Medium Imaging Surveys. By spatially cross-correlating photons in the GALEX far-UV (1500A) and near-UV (2300A) bands with a million spectroscopic objects in the Sloan Digital Sky Survey as a function of redshift, we robustly detect the redshift-dependent intensity of the UV background (UVB) modulated by its clustering bias up to $z\sim2$. These measurements clearly reveal the imprints of UVB spectral features redshifting through the filters. Using a simple parameterization, we simultaneously fit a UVB emissivity and clustering bias factor to these observations and constrain the main spectral features of the UVB spectrum: (i) the Lyman break, (ii) the non-ionizing UV continuum, which agrees with the Haardt & Madau model but does not rely on any assumption regarding the nature of the sources, and (iii) the Ly$α$ emission, whose luminosity density is consistent with estimates of the combined galaxy and AGN contributions at $z\sim1$. Because the technique probes the total background including low surface brightness emission, we place constraints on the amount of UV light originating from the diffuse intergalactic medium (IGM). Finally, the clustering bias of UV photons is found to be chromatic and evolving. Our frequency- and redshift-dependent UVB measurement delivers a summary statistic of the universe's net radiation output from stars, black holes, and the IGM combined.

astro-ph.CO↗