arXiv subjects

Risa H. Wechsler

Publications and source records attributed to Risa H. Wechsler.

At least 127 records · Page 7Linked to original sources

arXiv2018

Analyzing interferometric observations of strong gravitational lenses with recurrent and convolutional neural networks

We use convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to estimate the parameters of strong gravitational lenses from interferometric observations. We explore multiple strategies and find that the best results are obtained when the effects of the dirty beam are first removed from the images with a deconvolution performed with an RNN-based structure before estimating the parameters. For this purpose, we use the recurrent inference machine (RIM) introduced in Putzky & Welling (2017). This provides a fast and automated alternative to the traditional CLEAN algorithm. We obtain the uncertainties of the estimated parameters using variational inference with Bernoulli distributions. We test the performance of the networks with a simulated test dataset as well as with five ALMA observations of strong lenses. For the observed ALMA data we compare our estimates with values obtained from a maximum-likelihood lens modeling method which operates in the visibility space and find consistent results. We show that we can estimate the lensing parameters with high accuracy using a combination of an RNN structure performing image deconvolution and a CNN performing lensing analysis, with uncertainties less than a factor of two higher than those achieved with maximum-likelihood methods. Including the deconvolution procedure performed by RIM, a single evaluation can be done in about a second on a single GPU, providing a more than six orders of magnitude increase in analysis speed while using about eight orders of magnitude less computational resources compared to maximum-likelihood lens modeling in the uv-plane. We conclude that this is a promising method for the analysis of mm and cm interferometric data from current facilities (e.g., ALMA, JVLA) and future large interferometric observatories (e.g., SKA), where an analysis in the uv-plane could be difficult or unfeasible.

astro-ph.IM↗

arXiv2018

Hierarchical modeling and statistical calibration for photometric redshifts

The cosmological exploitation of modern photometric galaxy surveys requires both accurate (unbiased) and precise (narrow) redshift probability distributions derived from broadband photometry. Existing methodologies do not meet those requirements. Standard template fitting delivers interpretable models and errors, but lacks flexibility to learn inaccuracies in the observed photometry or the spectral templates. Machine learning addresses those issues, but requires representative training data, and the resulting models and uncertainties cannot be interpreted in the context of a physical model or outside of the training data. We present a hierarchical modeling approach simultaneously addressing the issues of flexibility, interpretability, and generalization. It combines template fitting with flexible (machine learning-like) models to correct the spectral templates, model their redshift distributions, and recalibrate the photometric observations. By optimizing the full posterior distribution of the model and solving for its (thousands of) parameters, one can perform a global statistical calibration of the data and the SED model. We apply this approach to the public Dark Energy Survey Science Verification data, and show that it provides more accurate and compact redshift posterior distributions than existing methods, as well as insights into residual photometric and SED systematics. The model is causal, makes predictions for future data (e.g., additional photometric bandpasses), and its internal parameters and components are interpretable. This approach does not formally require the training data to be complete or representative; in principle it can even work in regimes in which few or no spectroscopic redshifts are available.

astro-ph.CO↗

arXiv2018

Modeling the Impact of Baryons on Subhalo Populations with Machine Learning

We identify subhalos in dark matter-only (DMO) zoom-in simulations that are likely to be disrupted due to baryonic effects by using a random forest classifier trained on two hydrodynamic simulations of Milky Way (MW)-mass host halos from the Latte suite of the Feedback in Realistic Environments (FIRE) project. We train our classifier using five properties of each disrupted and surviving subhalo: pericentric distance and scale factor at first pericentric passage after accretion, and scale factor, virial mass, and maximum circular velocity at accretion. Our five-property classifier identifies disrupted subhalos in the FIRE simulations with an $85\%$ out-of-bag classification score. We predict surviving subhalo populations in DMO simulations of the FIRE host halos, finding excellent agreement with the hydrodynamic results; in particular, our classifier outperforms DMO zoom-in simulations that include the gravitational potential of the central galactic disk in each hydrodynamic simulation, indicating that it captures both the dynamical effects of a central disk and additional baryonic physics. We also predict surviving subhalo populations for a suite of DMO zoom-in simulations of MW-mass host halos, finding that baryons impact each system consistently and that the predicted amount of subhalo disruption is larger than the host-to-host scatter among the subhalo populations. Although the small size and specific baryonic physics prescription of our training set limits the generality of our results, our work suggests that machine-learning classification algorithms trained on hydrodynamic zoom-in simulations can efficiently predict realistic subhalo populations.

astro-ph.GA↗

arXiv2018

Mapping Extragalactic Dark Matter Annihilation with Galaxy Surveys: A Systematic Study of Stacked Group Searches

Dark matter in the halos surrounding galaxy groups and clusters can annihilate to high-energy photons. Recent advancements in the construction of galaxy group catalogs provide many thousands of potential extragalactic targets for dark matter. In this paper, we outline a procedure to infer the dark matter signal associated with a given galaxy group. Applying this procedure to a catalog of sources, one can create a full-sky map of the brightest extragalactic dark matter targets in the nearby Universe ($z\lesssim 0.03$), supplementing sources of dark matter annihilation from within the Local Group. As with searches for dark matter in dwarf galaxies, these extragalactic targets can be stacked together to enhance the signals associated with dark matter. We validate this procedure on mock $\textit{Fermi}$ gamma-ray data sets using a galaxy catalog constructed from the $\texttt{DarkSky}$ $N$-body cosmological simulation and demonstrate that the limits are robust, at $\mathcal{O}(1)$ levels, to systematic uncertainties on halo mass and concentration. We also quantify other sources of systematic uncertainty arising from the analysis and modeling assumptions. Our results suggest that a stacking analysis using galaxy group catalogs provides a powerful opportunity to discover extragalactic dark matter and complements existing studies of Milky Way dwarf galaxies.

astro-ph.CO↗

arXiv2018

The Aemulus Project III: Emulation of the Galaxy Correlation Function

Using the N-body simulations of the AEMULUS Project, we construct an emulator for the non-linear clustering of galaxies in real and redshift space. We construct our model of galaxy bias using the halo occupation framework, accounting for possible velocity bias. The model includes 15 parameters, including both cosmological and galaxy bias parameters. We demonstrate that our emulator achieves ~ 1% precision at the scales of interest, 0.1<r<10 h^{-1} Mpc, and recovers the true cosmology when tested against independent simulations. Our primary parameters of interest are related to the growth rate of structure, f, and its degenerate combination fsigma_8. Using this emulator, we show that the constraining power on these parameters monotonically increases as smaller scales are included in the analysis, all the way down to 0.1 h^{-1} Mpc. For a BOSS-like survey, the constraints on fsigma_8 from r<30 h^{-1} Mpc scales alone are more than a factor of two tighter than those from the fiducial BOSS analysis of redshift-space clustering using perturbation theory at larger scales. The combination of real- and redshift-space clustering allows us to break the degeneracy between f and sigma_8, yielding a 9% constraint on f alone for a BOSS-like analysis. The current AEMULUS simulations limit this model to surveys of massive galaxies. Future simulations will allow this framework to be extended to all galaxy target types, including emission-line galaxies.

astro-ph.CO↗

arXiv2018

The Aemulus Project II: Emulating the Halo Mass Function

Existing models for the dependence of the halo mass function on cosmological parameters will become a limiting source of systematic uncertainty for cluster cosmology in the near future. We present a halo mass function emulator and demonstrate improved accuracy relative to state-of-the-art analytic models. In this work, mass is defined using an overdensity criteria of 200 relative to the mean background density. Our emulator is constructed from the AEMULUS simulations, a suite of 40 N-body simulations with snapshots from z=3 to z=0. These simulations cover the flat wCDM parameter space allowed by recent Cosmic Microwave Background, Baryon Acoustic Oscillation and Type Ia Supernovae results, varying the parameters w, Omega_m, Omega_b, sigma_8, N_{eff}, n_s, and H_0. We validate our emulator using five realizations of seven different cosmologies, for a total of 35 test simulations. These test simulations were not used in constructing the emulator, and were run with fully independent initial conditions. We use our test simulations to characterize the modeling uncertainty of the emulator, and introduce a novel way of marginalizing over the associated systematic uncertainty. We confirm non-universality in our halo mass function emulator as a function of both cosmological parameters and redshift. Our emulator achieves better than 1% precision over much of the relevant parameter space, and we demonstrate that the systematic uncertainty in our emulator will remain a negligible source of error for cluster abundance studies through at least the LSST Year 1 data set.

astro-ph.CO↗

arXiv2018

The Aemulus Project I: Numerical Simulations for Precision Cosmology

The rapidly growing statistical precision of galaxy surveys has lead to a need for ever-more precise predictions of the observables used to constrain cosmological and galaxy formation models. The primary avenue through which such predictions will be obtained is suites of numerical simulations. These simulations must span the relevant model parameter spaces, be large enough to obtain the precision demanded by upcoming data, and be thoroughly validated in order to ensure accuracy. In this paper we present one such suite of simulations, forming the basis for the AEMULUS Project, a collaboration devoted to precision emulation of galaxy survey observables. We have run a set of 75 (1.05 h^-1 Gpc)^3 simulations with mass resolution and force softening of 3.51\times 10^10 (Omega_m / 0.3) ~ h^-1 M_sun and 20 ~ h^-1 kpc respectively in 47 different wCDM cosmologies spanning the range of parameter space allowed by the combination of recent Cosmic Microwave Background, Baryon Acoustic Oscillation and Type Ia Supernovae results. We present convergence tests of several observables including spherical overdensity halo mass functions, galaxy projected correlation functions, galaxy clustering in redshift space, and matter and halo correlation functions and power spectra. We show that these statistics are converged to 1% (2%) for halos with more than 500 (200) particles respectively and scales of r>200 ~ h^-1 kpc in real space or k ~ 3 h Mpc^-1 in harmonic space for z\le 1. We find that the dominant source of uncertainty comes from varying the particle loading of the simulations. This leads to large systematic errors for statistics using halos with fewer than 200 particles and scales smaller than k ~ 4 h^-1 Mpc. We provide the halo catalogs and snapshots detailed in this work to the community at https://AemulusProject.github.io.

astro-ph.CO↗

arXiv2018

The Connection between Galaxies and their Dark Matter Halos

In our modern understanding of galaxy formation, every galaxy forms within a dark matter halo. The formation and growth of galaxies over time is connected to the growth of the halos in which they form. The advent of large galaxy surveys as well as high-resolution cosmological simulations has provided a new window into the statistical relationship between galaxies and halos and its evolution. Here we define this galaxy-halo connection as the multi-variate distribution of galaxy and halo properties that can be derived from observations and simulations. This connection provides a key test of physical galaxy formation models; it also plays an essential role in constraints of cosmological models using galaxy surveys and in elucidating the properties of dark matter using galaxies. We review techniques for inferring the galaxy-halo connection and the insights that have arisen from these approaches. Some things we have learned are that galaxy formation efficiency is a strong function of halo mass; at its peak in halos around a pivot halo mass of 10^12 Msun, less than 20% of the available baryons have turned into stars by the present day; the intrinsic scatter in galaxy stellar mass is small, less than 0.2 dex at a given halo mass above this pivot mass; below this pivot mass galaxy stellar mass is a strong function of halo mass; the majority of stars over cosmic time were formed in a narrow region around this pivot mass. We also highlight key open questions about how galaxies and halos are connected, including understanding the correlations with secondary properties and the connection of these properties to galaxy clustering.

astro-ph.GA↗

arXiv2018

DESCQA: An Automated Validation Framework for Synthetic Sky Catalogs

The use of high-quality simulated sky catalogs is essential for the success of cosmological surveys. The catalogs have diverse applications, such as investigating signatures of fundamental physics in cosmological observables, understanding the effect of systematic uncertainties on measured signals and testing mitigation strategies for reducing these uncertainties, aiding analysis pipeline development and testing, and survey strategy optimization. The list of applications is growing with improvements in the quality of the catalogs and the details that they can provide. Given the importance of simulated catalogs, it is critical to provide rigorous validation protocols that enable both catalog providers and users to assess the quality of the catalogs in a straightforward and comprehensive way. For this purpose, we have developed the DESCQA framework for the Large Synoptic Survey Telescope Dark Energy Science Collaboration as well as for the broader community. The goal of DESCQA is to enable the inspection, validation, and comparison of an inhomogeneous set of synthetic catalogs via the provision of a common interface within an automated framework. In this paper, we present the design concept and first implementation of DESCQA. In order to establish and demonstrate its full functionality we use a set of interim catalogs and validation tests. We highlight several important aspects, both technical and scientific, that require thoughtful consideration when designing a validation framework, including validation metrics and how these metrics impose requirements on the synthetic sky catalogs.

astro-ph.IM↗

arXiv2018

Spatial Clustering of Dark Matter Halos: Secondary Bias, Neighbor Bias, and the Influence of Massive Neighbors on Halo Properties

We explore the phenomenon commonly known as halo assembly bias, whereby dark matter halos of the same mass are found to be more or less clustered when a second halo property is considered, for halos in the mass range $3.7 \times 10^{11} \; h^{-1} \mathrm{M_{\odot}} - 5.0 \times 10^{13} \; h^{-1} \mathrm{M_{\odot}}$. Using the Large Suite of Dark Matter Simulations (LasDamas) we consider nine commonly used halo properties and find that a clustering bias exists if halos are binned by mass or by any other halo property. This secondary bias implies that no single halo property encompasses all the spatial clustering information of the halo population. The mean values of some halo properties depend on their halo's distance to a more massive neighbor. Halo samples selected by having high values of one of these properties therefore inherit a neighbor bias such that they are much more likely to be close to a much more massive neighbor. This neighbor bias largely accounts for the secondary bias seen in halos binned by mass and split by concentration or age. However, halos binned by other mass-like properties still show a secondary bias even when the neighbor bias is removed. The secondary bias of halos selected by their spin behaves differently than that for other halo properties, suggesting that the origin of the spin bias is different than of other secondary biases.

astro-ph.CO↗

arXiv2017

The SAGA Survey: I. Satellite Galaxy Populations Around Eight Milky Way Analogs

We present the survey strategy and early results of the "Satellites Around Galactic Analogs" (SAGA) Survey. The SAGA Survey's goal is to measure the distribution of satellite galaxies around 100 systems analogous to the Milky Way down to the luminosity of the Leo I dwarf galaxy ($ M_r < -12.3 $). We define a Milky Way analog based on $K$-band luminosity and local environment. Here, we present satellite luminosity functions for 8 Milky Way analog galaxies between 20 to 40 Mpc. These systems have nearly complete spectroscopic coverage of candidate satellites within the projected host virial radius down to $ r_o < 20.75 $ using low redshift $gri$ color criteria. We have discovered a total of 25 new satellite galaxies: 14 new satellite galaxies meet our formal criteria around our complete host systems, plus 11 additional satellites in either incompletely surveyed hosts or below our formal magnitude limit. Combined with 13 previously known satellites, there are a total of 27 satellites around 8 complete Milky Way analog hosts. We find a wide distribution in the number of satellites per host, from 1 to 9, in the luminosity range for which there are five Milky Way satellites. Standard abundance matching extrapolated from higher luminosities predicts less scatter between hosts and a steeper luminosity function slope than observed. We find that the majority of satellites (26 of 27) are star-forming. These early results indicate that the Milky Way has a different satellite population than typical in our sample, potentially changing the physical interpretation of measurements based only on the Milky Way's satellite galaxies.

astro-ph.GA↗

arXiv2017

On estimation of contamination from hydrogen cyanide in carbon monoxide line intensity mapping

Line-intensity mapping surveys probe large-scale structure through spatial variations in molecular line emission from a population of unresolved cosmological sources. Future such surveys of carbon monoxide line emission, specifically the CO(1-0) line, face potential contamination from a disjoint population of sources emitting in a hydrogen cyanide emission line, HCN(1-0). This paper explores the potential range of the strength of HCN emission and its effect on the CO auto power spectrum, using simulations with an empirical model of the CO/HCN--halo connection. We find that effects on the observed CO power spectrum depend on modeling assumptions but are very small for our fiducial model based on our understanding of the galaxy--halo connection, with the bias in overall CO detection significance due to HCN expected to be less than 1%.

astro-ph.GA↗

arXiv2017

Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing

In Hezaveh et al. 2017 we showed that deep learning can be used for model parameter estimation and trained convolutional neural networks to determine the parameters of strong gravitational lensing systems. Here we demonstrate a method for obtaining the uncertainties of these parameters. We review the framework of variational inference to obtain approximate posteriors of Bayesian neural networks and apply it to a network trained to estimate the parameters of the Singular Isothermal Ellipsoid plus external shear and total flux magnification. We show that the method can capture the uncertainties due to different levels of noise in the input data, as well as training and architecture-related errors made by the network. To evaluate the accuracy of the resulting uncertainties, we calculate the coverage probabilities of marginalized distributions for each lensing parameter. By tuning a single hyperparameter, the dropout rate, we obtain coverage probabilities approximately equal to the confidence levels for which they were calculated, resulting in accurate and precise uncertainty estimates. Our results suggest that neural networks can be a fast alternative to Monte Carlo Markov Chains for parameter uncertainty estimation in many practical applications, allowing more than seven orders of magnitude improvement in speed.

astro-ph.CO↗

arXiv2017

Weak Lensing Measurement of the Mass--Richness Relation of SDSS redMaPPer Clusters

We perform a measurement of the mass--richness relation of the redMaPPer galaxy cluster catalogue using weak lensing data from the Sloan Digital Sky Survey. We have carefully characterized a broad range of systematic uncertainties, including shear calibration errors, photo-$z$ biases, dilution by member galaxies, source obscuration, magnification bias, incorrect assumptions about cluster mass profiles, cluster centering, halo triaxiality, and projection effects. We also compare measurements of the lensing signal from two independently-produced shear and photometric redshift catalogues to characterize systematic errors in the lensing signal itself. Using a sample of 5,570 clusters from $0.1\le z\le 0.33$, the normalization of our power-law mass vs.\ $λ$ relation is $\log_{10}[M_{200m}/h^{-1}\ M_{\odot}]$ = $14.344 \pm 0.021$ (statistical) $\pm 0.023$ (systematic) at a richness $λ=40$, a 7 per cent calibration uncertainty, with a power-law index of $1.33^{+0.09}_{-0.10}$ ($1σ$). The detailed systematics characterization in this work renders it the definitive weak lensing mass calibration for SDSS redMaPPer clusters at this time.

astro-ph.CO↗

arXiv2016

The Concentration Dependence of the Galaxy-Halo Connection: Modeling Assembly Bias with Abundance Matching

Empirical methods for connecting galaxies to their dark matter halos have become essential for interpreting measurements of the spatial statistics of galaxies. In this work, we present a novel approach for parameterizing the degree of concentration dependence in the abundance matching method. This new parameterization provides a smooth interpolation between two commonly used matching proxies: the peak halo mass and the peak halo maximal circular velocity. This parameterization controls the amount of dependence of galaxy luminosity on halo concentration at a fixed halo mass. Effectively this interpolation scheme enables abundance matching models to have adjustable assembly bias in the resulting galaxy catalogs. With the new 400 Mpc/h DarkSky Simulation, whose larger volume provides lower sample variance, we further show that low-redshift two-point clustering and satellite fraction measurements from SDSS can already provide a joint constraint on this concentration dependence and the scatter within the abundance matching framework.

astro-ph.CO↗

arXiv2016

The Faber-Jackson relation and Fundamental Plane from halo abundance matching

The Fundamental Plane (FP) describes the relation between the stellar mass, size, and velocity dispersion of elliptical galaxies; the Faber-Jackson relation (FJR) is its projection onto {mass, velocity} space. In this work we redeploy and expand the framework of Desmond & Wechsler (2015) to ask whether abundance matching-based LCDM models that have shown success in matching the spatial distribution of galaxies are also capable of explaining key properties of the FJR and FP, including their scatter. Within our framework, agreement with the normalisation of the FJR requires haloes to expand in response to disc formation. We find that the tilt of the FP may be explained by a combination of the observed non-homology in galaxy structure and the variation in mass-to-light ratio produced by abundance matching with a universal initial mass function (IMF), provided that the anisotropy of stellar motions is taken into account. However, the predicted scatter around the FP is considerably increased by situating galaxies in cosmologically-motivated haloes due to variations in halo properties at fixed stellar mass, and appears to exceed that of the data. This implies that additional correlations between galaxy and halo variables may be required to fully reconcile these models with elliptical galaxy scaling relations.

astro-ph.GA↗

arXiv2016

EmpiriciSN: Re-sampling Observed Supernova/Host Galaxy Populations using an XD Gaussian Mixture Model

We describe two new open source tools written in Python for performing extreme deconvolution Gaussian mixture modeling (XDGMM) and using a conditioned model to re-sample observed supernova and host galaxy populations. XDGMM is new program for using Gaussian mixtures to do density estimation of noisy data using extreme deconvolution (XD) algorithms that has functionality not available in other XD tools. It allows the user to select between the AstroML (Vanderplas et al. 2012; Ivezic et al. 2015) and Bovy et al. (2011) fitting methods and is compatible with scikit-learn machine learning algorithms (Pedregosa et al. 2011). Most crucially, it allows the user to condition a model based on the known values of a subset of parameters. This gives the user the ability to produce a tool that can predict unknown parameters based on a model conditioned on known values of other parameters. EmpiriciSN is an example application of this functionality that can be used for fitting an XDGMM model to observed supernova/host datasets and predicting likely supernova parameters using on a model conditioned on observed host properties. It is primarily intended for simulating realistic supernovae for LSST data simulations based on empirical galaxy properties.

astro-ph.IM↗

arXiv2016

Biased Tracers in Redshift Space in the EFT of Large-Scale Structure

The Effective Field Theory of Large-Scale Structure (EFTofLSS) provides a novel formalism that is able to accurately predict the clustering of large-scale structure (LSS) in the mildly non-linear regime. Here we provide the first computation of the power spectrum of biased tracers in redshift space at one loop order, and we make the associated code publicly available. We compare the multipoles $\ell=0,2$ of the redshift-space halo power spectrum, together with the real-space matter and halo power spectra, with data from numerical simulations at $z=0.67$. For the samples we compare to, which have a number density of $\bar n=3.8 \cdot 10^{-2}(h \ {\rm Mpc}^{-1})^3$ and $\bar n=3.9 \cdot 10^{-4}(h \ {\rm Mpc}^{-1})^3$, we find that the calculation at one-loop order matches numerical measurements to within a few percent up to $k\simeq 0.43 \ h \ {\rm Mpc}^{-1}$, a significant improvement with respect to former techniques. By performing the so-called IR-resummation, we find that the Baryon Acoustic Oscillation peak is accurately reproduced. Based on the results presented here, long-wavelength statistics that are routinely observed in LSS surveys can be finally computed in the EFTofLSS. This formalism thus is ready to start to be compared directly to observational data.

astro-ph.CO↗