arXiv subjects

Peter Melchior

Publications and source records attributed to Peter Melchior.

At least 37 records · Page 2Linked to original sources

arXiv2024

The little coadd that could: Estimating shear from coadded images

Upcoming wide field surveys will have many overlapping epochs of the same region of sky. The conventional wisdom is that in order to reduce the errors sufficiently for systematics-limited measurements, like weak lensing, we must do simultaneous fitting of all the epochs. Using current algorithms this will require a significant amount of computing time and effort. In this paper, we revisit the potential of using coadds for shear measurements. We show on a set of image simulations that the multiplicative shear bias can be constrained below the 0.1% level on coadds, which is sufficient for future lensing surveys. We see no significant differences between simultaneous fitting and coadded approaches for two independent shear codes: Metacalibration and BFD. One caveat of our approach is the assumption of a principled coadd, i.e. the PSF is mathematically well-defined for all the input images. This requires us to reject CCD images that do not fully cover the coadd region. We estimate that the number of epochs that must be rejected for a survey like LSST is on the order of 20%, resulting in a small loss in depth of less than 0.1 magnitudes. We also put forward a cell-based coaddition scheme that meets the above requirements for unbiased weak lensing shear estimation in the context of LSST.

astro-ph.CO↗

arXiv2024

Constraining Protoplanetary Disk Winds from Forbidden Line Profiles with Simulation-based Inference

Protoplanetary disks are the sites of vigorous hydrodynamic processes, such as accretion and outflows, and ultimately establish the conditions for the formation of planets. The properties of disk outflows are often inferred through analysis of forbidden emission lines. These lines contain multiple overlapping components, tracing different emission regions with different processes that excite them: a high-velocity component (tracing a jet), a broad low-velocity component (tracing inner disk wind), and a narrow low-velocity component (tracing outer disk wind). They are also heavily contaminated by background spectral features. All of these challenges call into question the traditional approach of fitting Gaussian components to the line profiles, and cloud the physical interpretation of those components. We introduce a novel statistical technique to analyze emission lines in protoplanetary disks. Simulation-Based Inference is a computationally efficient machine learning technique that produces posterior distributions of the parameters (e.g. magnetic field, radiation sources, geometry) of a representative wind model when given a spectrum, without any prior assumption about line shapes (e.g.symmetry). In this pathfinder study, we demonstrate that this technique indeed accurately recovers the parameters from simulated spectra without noise and background. A following work will deal with the analysis of observed spectra.

astro-ph.SR↗

arXiv2023

PopSED: Population-Level Inference for Galaxy Properties from Broadband Photometry with Neural Density Estimation

We present PopSED, a framework for the population-level inference of galaxy properties from photometric data. Unlike the traditional approach of first analyzing individual galaxies and then combining the results to determine the physical properties of the entire galaxy population, we directly make the population distribution the inference objective. We train normalizing flows to approximate the population distribution by minimizing the Wasserstein distance between the synthetic photometry of the galaxy population and the observed data. We validate our method using mock observations and apply it to galaxies from the GAMA survey. PopSED reliably recovers the redshift and stellar mass distribution of $10^{5}$ galaxies using broadband photometry within $<1$ GPU hr, being $10^{5-6}$ times faster than the traditional spectral energy distribution modeling method. From the population posterior, we also recover the star-forming main sequence for GAMA galaxies at $z<0.1$. With the unprecedented number of galaxies in upcoming surveys, our method offers an efficient tool for studying galaxy evolution and deriving redshift distributions for cosmological analyses.

astro-ph.IM↗

arXiv2023

AESTRA: Deep Learning for Precise Radial Velocity Estimation in the Presence of Stellar Activity

Stellar activity interferes with precise radial velocity measurements and limits our ability to detect and characterize planets, particularly Earth-like planets. We introduce \aestra (Auto-Encoding STellar Radial-velocity and Activity), a deep learning method for precise radial velocity measurements. It combines a spectrum auto-encoder, which learns to create realistic models of the star's rest-frame spectrum, and a radial-velocity estimator, which learns to identify true Doppler shifts in the presence of spurious shifts due to line-profile variations. Being self-supervised, \aestra does not need "ground truth" radial velocities for training, making it applicable to exoplanet host stars for which the truth is unknown. In tests involving 1,000 simulated spectra, \aestra can detect planetary signals as low as 0.1 m/s even in the presence of 3 m/s of activity-induced noise and 0.3 m/s of photon noise per spectrum.

astro-ph.EP↗

arXiv2023

Joint Cosmic Density Reconstruction from Photometric and Spectroscopic Samples

We reconstruct the dark matter density field from spatially overlapping spectroscopic and photometric redshift catalogs through a forward modelling approach. Instead of directly inferring the underlying density field, we find the best fitting initial Gaussian fluctuations that will evolve into the observed cosmic volume. To account for the substantial uncertainty of photometric redshifts we employ a differentiable continuous Poisson process. In the context of the upcoming Prime Focus Spectrograph (PFS), we find improvements in cosmic structure classification equivalent to 50-100\% more spectroscopic targets by combining relatively sparse spectroscopic with dense photometric samples.

astro-ph.CO↗

arXiv2023

Multiscale Feature Attribution for Outliers

Machine learning techniques can automatically identify outliers in massive datasets, much faster and more reproducible than human inspection ever could. But finding such outliers immediately leads to the question: which features render this input anomalous? We propose a new feature attribution method, Inverse Multiscale Occlusion, that is specifically designed for outliers, for which we have little knowledge of the type of features we want to identify and expect that the model performance is questionable because anomalous test data likely exceed the limits of the training data. We demonstrate our method on outliers detected in galaxy spectra from the Dark Energy Survey Instrument and find its results to be much more interpretable than alternative attribution approaches.

cs.LG↗

arXiv2023

Cosmology with Galaxy Photometry Alone

We present the first cosmological constraints using only the observed photometry of galaxies. Villaescusa-Navarro et al. (2022; arXiv:2201.02202) recently demonstrated that the internal physical properties of a single simulated galaxy contain a significant amount of cosmological information. These physical properties, however, cannot be directly measured from observations. In this work, we present how we can go beyond theoretical demonstrations to infer cosmological constraints from actual galaxy observables (e.g. optical photometry) using neural density estimation and the CAMELS suite of hydrodynamical simulations. We find that the cosmological information in the photometry of a single galaxy is limited. However, we combine the constraining power of photometry from many galaxies using hierarchical population inference and place significant cosmological constraints. With the observed photometry of $\sim$20,000 NASA-Sloan Atlas galaxies, we constrain $Ω_m = 0.323^{+0.075}_{-0.095}$ and $σ_8 = 0.799^{+0.088}_{-0.085}$.

astro-ph.CO↗

arXiv2023

Beyond Ultra-diffuse Galaxies. I. Mass--Size Outliers among the Satellites of Milky Way Analogs

Large diffuse galaxies are hard to find, but understanding the environments where they live, their numbers, and ultimately their origins, is of intense interest and importance for galaxy formation and evolution. Using Subaru's Hyper Suprime-Cam Strategic Survey Program, we perform a systematic search for low surface brightness galaxies and present novel and effective methods for detecting and modeling them. As a case study, we surveyed 922 Milky Way analogs in the nearby Universe ($0.01 < z < 0.04$) and build a large sample of satellite galaxies that are outliers in the mass-size relation. These ``ultra-puffy'' galaxies (UPGs), defined to be $1.5σ$ above the average mass-size relation, represent the tail of the satellite size distribution. We find that each MW analog hosts $N_{\rm UPG} = 0.31\pm 0.05$ ultra-puffy galaxies on average, which is consistent with but slightly lower than the observed abundance at this halo mass in the Local Volume. We also construct a sample of ultra-diffuse galaxies (UDGs) in MW analogs and find an abundance of $N_{\rm UDG} = 0.44\pm0.05$ per host. With literature results, we confirm that the UDG abundance scales with the host halo mass following a sublinear power law. We argue that our definition for ultra-puffy galaxies, which is based on the mass-size relation, is more physically-motivated than the common definition of ultra-diffuse galaxies, which depends on surface brightness and size cuts and thus yields different surface mass density cuts for quenched and star-forming galaxies.

astro-ph.GA↗

arXiv2023

Outlier Detection in the DESI Bright Galaxy Survey

We present an unsupervised search for outliers in the Bright Galaxy Survey (BGS) dataset from the DESI Early Data Release. This analysis utilizes an autoencoder to compress galaxy spectra into a compact, redshift-invariant latent space, and a normalizing flow to identify low-probability objects. The most prominent outliers show distinctive spectral features such as irregular or double-peaked emission lines, or originate from galaxy mergers, blended sources, and rare quasar types, including one previously unknown Broad Absorption Line system. A significant portion of the BGS outliers are stars spectroscopically misclassified as galaxies. By building our own star model trained on spectra from the DESI Milky Way Survey, we have determined that the misclassification likely stems from the Principle Component Analysis of stars in the DESI pipeline. To aid follow-up studies, we make the full probability catalog of all BGS objects and our pre-trained models publicly available.

astro-ph.GA↗

arXiv2023

Spotting Hallucinations in Inverse Problems with Data-Driven Priors

Hallucinations are an inescapable consequence of solving inverse problems with deep neural networks. The expressiveness of recent generative models is the reason why they can yield results far superior to conventional regularizers; it can also lead to realistic-looking but incorrect features, potentially undermining the trust in important aspects of the reconstruction. We present a practical and computationally efficient method to determine, which regions in the solutions of inverse problems with data-driven priors are prone to hallucinations. By computing the diagonal elements of the Fisher information matrix of the likelihood and the data-driven prior separately, we can flag regions where the information is prior-dominated. Our diagnostic can directly be compared to the reconstructed solutions and enables users to decide if measurements in such regions are robust for their application. Our method scales linearly with the number of parameters and is thus applicable in high-dimensional settings, allowing it to be rolled out broadly for the large-volume data products of future wide-field surveys.

astro-ph.IM↗

arXiv2023

Autoencoding Galaxy Spectra II: Redshift Invariance and Outlier Detection

We present an unsupervised outlier detection method for galaxy spectra based on the spectrum autoencoder architecture spender, which reliably captures spectral features and provides highly realistic reconstructions for SDSS galaxy spectra. We interpret the sample density in the autoencoder latent space as a probability distribution, and identify outliers as low-probability objects with a normalizing flow. However, we found that the latent-space position is not, as expected from the architecture, redshift invariant, which introduces stochasticity into the latent space and the outlier detection method. We solve this problem by adding two novel loss terms during training, which explicitly link latent-space distances to data-space distances, preserving locality in the autoencoding process. Minimizing the additional losses leads to a redshift-invariant, non-degenerate latent space distribution with clear separations between common and anomalous data. We inspect the spectra with the lowest probability and find them to include blends with foreground stars, extremely reddened galaxies, galaxy pairs and triples, and stars that are misclassified as galaxies. We release the newly trained spender model and the latent-space probability for the entire SDSS-I galaxy sample to aid further investigations.

astro-ph.IM↗

arXiv2022

The Third Gravitational Lensing Accuracy Testing (GREAT3) Challenge Handbook

The GRavitational lEnsing Accuracy Testing 3 (GREAT3) challenge is the third in a series of image analysis challenges, with a goal of testing and facilitating the development of methods for analyzing astronomical images that will be used to measure weak gravitational lensing. This measurement requires extremely precise estimation of very small galaxy shape distortions, in the presence of far larger intrinsic galaxy shapes and distortions due to the blurring kernel caused by the atmosphere, telescope optics, and instrumental effects. The GREAT3 challenge is posed to the astronomy, machine learning, and statistics communities, and includes tests of three specific effects that are of immediate relevance to upcoming weak lensing surveys, two of which have never been tested in a community challenge before. These effects include realistically complex galaxy models based on high-resolution imaging from space; spatially varying, physically-motivated blurring kernel; and combination of multiple different exposures. To facilitate entry by people new to the field, and for use as a diagnostic tool, the simulation software for the challenge is publicly available, though the exact parameters used for the challenge are blinded. Sample scripts to analyze the challenge data using existing methods will also be provided. See http://great3challenge.info and http://great3.projects.phys.ucl.ac.uk/leaderboard/ for more information.

astro-ph.CO↗

arXiv2022

Plausible Adversarial Attacks on Direct Parameter Inference Models in Astrophysics

In this abstract we explore the possibility of introducing biases in physical parameter inference models from adversarial-type attacks. In particular, we inject small amplitude systematics into inputs to a mixture density networks tasked with inferring cosmological parameters from observed data. The systematics are constructed analogously to white-box adversarial attacks. We find that the analysis network can be tricked into spurious detection of new physics in cases where standard cosmological estimators would be insensitive. This calls into question the robustness of such networks and their utility for reliably detecting new physics.

astro-ph.CO↗

arXiv2022

Autoencoding Galaxy Spectra I: Architecture

We introduce the neural network architecture SPENDER as a core differentiable building block for analyzing, representing, and creating galaxy spectra. It combines a convolutional encoder, which pays attention to up to 256 spectral features and compresses them into a low-dimensional latent space, with a decoder that generates a restframe representation, whose spectral range and resolution exceeds that of the observing instrument. The decoder is followed by explicit redshift, resampling, and convolution transformations to match the observations. The architecture takes galaxy spectra at arbitrary redshifts and is robust to glitches like residuals of the skyline subtraction, so that spectra from a large survey can be ingested directly without additional preprocessing. We demonstrate the performance of SPENDER by training on the entire spectroscopic galaxy sample of SDSS-II; show its ability to create highly accurate reconstructions with substantially reduced noise; perform deconvolution and oversampling for a super-resolution model that resolves the [OII] doublet; introduce a novel method to interpret attention weights as proxies for important spectral features; and infer the main degrees of freedom represented in the latent space. We conclude with a discussion of future improvements and applications.

astro-ph.IM↗

arXiv2022

Deblending Galaxies with Generative Adversarial Networks

Deep generative models including generative adversarial networks (GANs) are powerful unsupervised tools in learning the distributions of data sets. Building a simple GAN architecture in PyTorch and training on the CANDELS data set, we generate galaxy images with the Hubble Space Telescope resolution starting from a noise vector. We proceed by modifying the GAN architecture to improve the Subaru Hyper Suprime-Cam ground-based images by increasing their resolution to the HST resolution. We use the super resolution GAN on a large sample of blended galaxies which we create using CANDELS cutouts. In our simulated blend sample, $\sim 20 \%$ would unrecognizably be blended even in the HST resolution cutouts. In the HSC-like cutouts this fraction rises to $\sim 90\%$. With our modified GAN we can lower this value to $\sim 50\%$. We quantify the blending fraction in the high, low and GAN resolutions over the whole manifold of angular separation, flux ratios, sizes and redshift difference between the two blended objects. The two peaks found by the GAN deblender result in ten times improvement in the photometry measurement of the blended objects. Modifying the architecture of the GAN, we also train a Multi-wavelength GAN with seven band optical+NIR HST cutouts. This multi-wavelength GAN improves the fraction of detected blends by another $\sim 10\%$ compared to the single-band GAN. This is most beneficial to the current and future precision cosmology experiments (e.g., LSST, SPHEREx, Euclid, Roman), specifically those relying on weak gravitational lensing, where blending is a major source of systematic error.

astro-ph.CO↗

arXiv2022

$\texttt{Mangrove}$: Learning Galaxy Properties from Merger Trees

Efficiently mapping baryonic properties onto dark matter is a major challenge in astrophysics. Although semi-analytic models (SAMs) and hydrodynamical simulations have made impressive advances in reproducing galaxy observables across cosmologically significant volumes, these methods still require significant computation times, representing a barrier to many applications. Graph Neural Networks (GNNs) have recently proven to be the natural choice for learning physical relations. Among the most inherently graph-like structures found in astrophysics are the dark matter merger trees that encode the evolution of dark matter halos. In this paper we introduce a new, graph-based emulator framework, $\texttt{Mangrove}$, and show that it emulates the galactic stellar mass, cold gas mass and metallicity, instantaneous and time-averaged star formation rate, and black hole mass -- as predicted by a SAM -- with root mean squared error up to two times lower than other methods across a $(75 Mpc/h)^3$ simulation box in 40 seconds, 4 orders of magnitude faster than the SAM. We show that $\texttt{Mangrove}$ allows for quantification of the dependence of galaxy properties on merger history. We compare our results to the current state of the art in the field and show significant improvements for all target properties. $\texttt{Mangrove}$ is publicly available.

astro-ph.GA↗

arXiv2022

Rubin-Euclid Derived Data Products: Initial Recommendations

This report is the result of a joint discussion between the Rubin and Euclid scientific communities. The work presented in this report was focused on designing and recommending an initial set of Derived Data products (DDPs) that could realize the science goals enabled by joint processing. All interested Rubin and Euclid data rights holders were invited to contribute via an online discussion forum and a series of virtual meetings. Strong interest in enhancing science with joint DDPs emerged from across a wide range of astrophysical domains: Solar System, the Galaxy, the Local Volume, from the nearby to the primaeval Universe, and cosmology.

astro-ph.IM↗

arXiv2022

Lightweight starshade position sensing with convolutional neural networks and simulation-based inference

Starshades are a leading technology to enable the direct detection and spectroscopic characterization of Earth-like exoplanets. To keep the starshade and telescope aligned over large separations, reliable sensing of the peak of the diffracted light of the occluded star is required. Current techniques rely on image matching or model fitting, both of which put substantial computational burdens on resource-limited spacecraft computers. We present a lightweight image processing method based on a convolutional neural network paired with a simulation-based inference technique to estimate the position of the spot of Arago and its uncertainty. The method achieves an accuracy of a few centimeters across the entire pupil plane, while only requiring 1.6 MB in stored data structures and 5.3 MFLOPs (million floating point operations) per image at test time. By deploying our method at the Princeton Starshade Testbed, we demonstrate that the neural network can be trained on simulated images and used on real images, and that it can successfully be integrated in the control system for closed-loop formation flying.

astro-ph.IM↗