arXiv subjects

Arya Farahi

Publications and source records attributed to Arya Farahi.

At least 55 records · Page 3Linked to original sources

arXiv2021

SHAPing the Gas: Understanding Gas Shapes in Dark Matter Haloes with Interpretable Machine Learning

The non-spherical shapes of dark matter and gas distributions introduce systematic uncertainties that affect observable-mass relations and selection functions of galaxy groups and clusters. However, the triaxial gas distributions depend on the non-linear physical processes of halo formation histories and baryonic physics, which are challenging to model accurately. In this study we explore a machine learning approach for modelling the dependence of gas shapes on dark matter and baryonic properties. With data from the IllustrisTNG hydrodynamical cosmological simulations, we develop a machine learning pipeline that applies \pkg{XGBoost}, an implementation of gradient boosted decision trees, to predict radial profiles of gas shapes from halo properties. We show that \pkg{XGBoost} models can accurately predict gas shape profiles in dark matter haloes. We also explore model interpretability with \pkg{SHAP}, a method that identifies the most predictive properties at different halo radii. We find that baryonic properties best predict gas shapes in halo cores, whereas dark matter shapes are the main predictors in the halo outskirts. This work demonstrates the power of interpretable machine learning in modelling observable properties of dark matter haloes in the era of multi-wavelength cosmological surveys.

astro-ph.CO↗

arXiv2021

Approximate Bayesian Uncertainties on Deep Learning Dynamical Mass Estimates of Galaxy Clusters

We study methods for reconstructing Bayesian uncertainties on dynamical mass estimates of galaxy clusters using convolutional neural networks (CNNs). We discuss the statistical background of approximate Bayesian neural networks and demonstrate how variational inference techniques can be used to perform computationally tractable posterior estimation for a variety of deep neural architectures. We explore how various model designs and statistical assumptions impact prediction accuracy and uncertainty reconstruction in the context of cluster mass estimation. We measure the quality of our model posterior recovery using a mock cluster observation catalog derived from the MultiDark simulation and UniverseMachine catalog. We show that approximate Bayesian CNNs produce highly accurate dynamical cluster mass posteriors. These model posteriors are log-normal in cluster mass and recover $68\%$ and $90\%$ confidence intervals to within $1\%$ of their measured value. We note how this rigorous modeling of dynamical mass posteriors is necessary for using cluster abundance measurements to constrain cosmological parameters.

astro-ph.CO↗

arXiv2021

The Role of Machine Learning in the Next Decade of Cosmology

In recent years, machine learning (ML) methods have remarkably improved how cosmologists can interpret data. The next decade will bring new opportunities for data-driven cosmological discovery, but will also present new challenges for adopting ML methodologies and understanding the results. ML could transform our field, but this transformation will require the astronomy community to both foster and promote interdisciplinary research endeavors.

astro-ph.IM↗

arXiv2020

A Robust and Efficient Deep Learning Method for Dynamical Mass Measurements of Galaxy Clusters

We demonstrate the ability of convolutional neural networks (CNNs) to mitigate systematics in the virial scaling relation and produce dynamical mass estimates of galaxy clusters with remarkably low bias and scatter. We present two models, CNN$_\mathrm{1D}$ and CNN$_\mathrm{2D}$, which leverage this deep learning tool to infer cluster masses from distributions of member galaxy dynamics. Our first model, CNN$_\text{1D}$, infers cluster mass directly from the distribution of member galaxy line-of-sight velocities. Our second model, CNN$_\text{2D}$, extends the input space of CNN$_\text{1D}$ to learn on the joint distribution of galaxy line-of-sight velocities and projected radial distances. We train each model as a regression over cluster mass using a labeled catalog of realistic mock cluster observations generated from the MultiDark simulation and UniverseMachine catalog. We then evaluate the performance of each model on an independent set of mock observations selected from the same simulated catalog. The CNN models produce cluster mass predictions with lognormal residuals of scatter as low as $0.132$ dex, greater than a factor of 2 improvement over the classical $M$-$σ$ power-law estimator. Furthermore, the CNN model reduces prediction scatter relative to similar machine learning approaches by up to $17\%$ while executing in drastically shorter training and evaluation times (by a factor of 30) and producing considerably more robust mass predictions (improving prediction stability under variations in galaxy sampling rate by $30\%$).

astro-ph.CO↗

arXiv2020

Driving with Data in the Motor City: Mining and Modeling Vehicle Fleet Maintenance Data

The City of Detroit maintains an active fleet of over 2500 vehicles, spending an annual average of over \$5 million on purchases and over \$7.7 million on maintenance. Modeling patterns and trends in this data is of particular importance to a variety of stakeholders, particularly as Detroit emerges from Chapter 9 bankruptcy, but the structure in such data is complex, and the city lacks dedicated resources for in-depth analysis. The City of Detroit's Operations and Infrastructure Group and the University of Michigan initiated a collaboration which seeks to address this unmet need by analyzing data from the City of Detroit's vehicle fleet. This work presents a case study and provides the first data-driven benchmark, demonstrating a suite of methods to aid in data understanding and prediction for large vehicle maintenance datasets. We present analyses to address three key questions raised by the stakeholders, related to discovering multivariate maintenance patterns over time; predicting maintenance; and predicting vehicle- and fleet-level costs. We present a novel algorithm, PRISM, for automating multivariate sequential data analyses using tensor decomposition. This work is a first of its kind that presents both methodologies and insights to guide future civic data research.

cs.CY↗

arXiv2020

PoPE: A population-based approach to model spatial structure of astronomical systems

We present a novel population-based Bayesian inference approach to model the average and population variance of spatial distribution of a set of observables from ensemble analysis of low signal-to-noise ratio measurements. The method consists of (1) inferring the average profile using Gaussian Processes and (2) computing the covariance of the profile observables given a set of independent variables. Our model is computationally efficient and capable of inferring average profiles of a large population size from noisy measurements, without stacking and binning data nor parameterizing the shape of the mean profile. We demonstrate the performance of our method using dark matter, gas and stellar profiles extracted from hydrodynamical cosmological simulations of galaxy formation. Population Profile Estimator (PoPE) is publicly available in a GitHub repository. Our new method should be useful for measuring the spatial distribution and internal structure of a variety of astrophysical systems using large astronomical surveys.

astro-ph.CO↗

arXiv2020

Dark Energy Survey Year 1 Results: Cosmological Constraints from Cluster Abundances and Weak Lensing

We perform a joint analysis of the counts and weak lensing signal of redMaPPer clusters selected from the Dark Energy Survey (DES) Year 1 dataset. Our analysis uses the same shear and source photometric redshifts estimates as were used in the DES combined probes analysis. Our analysis results in surprisingly low values for $S_8 =σ_8(Ω_{\rm m}/0.3)^{0.5}= 0.65\pm 0.04$, driven by a low matter density parameter, $Ω_{\rm m}=0.179^{+0.031}_{-0.038}$, with $σ_8-Ω_{\rm m}$ posteriors in $2.4σ$ tension with the DES Y1 3x2pt results, and in $5.6σ$ with the Planck CMB analysis. These results include the impact of post-unblinding changes to the analysis, which did not improve the level of consistency with other data sets compared to the results obtained at the unblinding. The fact that multiple cosmological probes (supernovae, baryon acoustic oscillations, cosmic shear, galaxy clustering and CMB anisotropies), and other galaxy cluster analyses all favor significantly higher matter densities suggests the presence of systematic errors in the data or an incomplete modeling of the relevant physics. Cross checks with X-ray and microwave data, as well as independent constraints on the observable--mass relation from SZ selected clusters, suggest that the discrepancy resides in our modeling of the weak lensing signal rather than the cluster abundance. Repeating our analysis using a higher richness threshold ($λ\ge 30$) significantly reduces the tension with other probes, and points to one or more richness-dependent effects not captured by our model.

astro-ph.CO↗

arXiv2020

Aging Halos: Implications of the Magnitude Gap on Conditional Statistics of Stellar and Gas Properties of Massive Halos

Cold dark matter model predicts that the large-scale structure grows hierarchically. Small dark matter halos form first. Then, they grow gradually via continuous merger and accretion. These halos host the majority of baryonic matter in the Universe in the form of hot gas and cold stellar phase. Determining how baryons are partitioned into these phases requires detailed modeling of galaxy formation and their assembly history. It is speculated that formation time of the same mass halos might be correlated with their baryonic content. To evaluate this hypothesis, we employ halos of mass above $10^{14}\,M_{\odot}$ realized by TNG300 solution of the IllustrisTNG project. Formation time is not directly observable. Hence, we rely on the magnitude gap between the brightest and the fourth brightest halo galaxy member, which is shown that traces formation time of the host halo. We compute the conditional statistics of the stellar and gas content of halos conditioned on their total mass and magnitude gap. We find a strong correlation between magnitude gap and gas mass, BCG stellar mass, and satellite galaxies stellar mass, but not the total stellar mass of halo. Conditioning on the magnitude gap can reduce the scatter about halo property--halo mass relation and has a significant impact on the conditional covariance. Reduction in the scatter can be as significant as 30%, which implies more accurate halo mass prediction. Incorporating the magnitude gap has a potential to improve cosmological constraints using halo abundance and allows us to gain insight into the baryon evolution within these systems.

astro-ph.GA↗

arXiv2020

Stellar Property Statistics of Massive Halos from Cosmological Hydrodynamics Simulations: Common Kernel Shapes

We study stellar property statistics, including satellite galaxy occupation, of massive halo populations realized by three cosmological hydrodynamics simulations: BAHAMAS + MACSIS, TNG300 of the IllustrisTNG suite, and Magneticum Pathfinder. The simulations incorporate independent sub-grid methods for astrophysical processes with spatial resolutions ranging from $1.5$ to $6$ kpc, and each generates samples of $1000$ or more halos with $M_{\rm halo}> 10^{13.5} M_{\odot}$ at redshift $z=0$. Applying localized, linear regression (LLR), we extract halo mass-conditioned statistics (normalizations, slopes, and intrinsic covariance) for a three-element stellar property vector consisting of: i) $N_{sat}$, the number of satellite galaxies with stellar mass, $M_{\star, \rm sat} > 10^{10} M_{\odot}$ within radius $R_{200c}$ of the halo; ii) $M_{\star,\rm tot}$, the total stellar mass within that radius, and; iii) $M_{\star,\rm BCG}$, the gravitationally-bound stellar mass of the central galaxy within a $100 \, \rm kpc$ radius. Scaling parameters for the three properties with halo mass show mild differences among the simulations, in part due to numerical resolution, but there is qualitative agreement on property correlations, with halos having smaller than average central galaxies tending to also have smaller total stellar mass and a larger number of satellite galaxies. Marginalizing over total halo mass, we find the satellite galaxy kernel, $p(\ln N_{sat}\,|\,M_{\rm halo},z)$ to be consistently skewed left, with skewness parameter $γ= -0.91 \pm 0.02$, while that of $\ln M_{\star,\rm tot}$ is closer to log-normal, in all three simulations. The highest resolution simulations find $γ\simeq -0.8$ for the $z=0$ shape of $p(\ln M_{\star,\rm BCG}\,|\,M_{\rm halo},z)$ and also that the fractional scatter in total stellar mass is below $10\%$ in halos more massive than $10^{14.3} M_{\odot}$.

astro-ph.GA↗

arXiv2019

Chandra Follow-Up of the SDSS DR8 redMaPPer Catalog Using the MATCha Pipeline

In order to place constraints on cosmology through optical surveys of galaxy clusters, one must first understand the properties of those clusters. To this end, we introduce the Mass Analysis Tool for Chandra (MATCha), a pipeline which uses a parallellized algorithm to analyze archival Chandra data. MATCha simultaneously calculates X-ray temperatures and luminosities and performs centering measurements for hundreds of potential galaxy clusters using archival X-ray exposures. We run MATCha on the redMaPPer SDSS DR8 cluster catalog and use MATCha's output X-ray temperatures and luminosities to analyze the galaxy cluster temperature-richness, luminosity-richness, luminosity-temperature, and temperature-luminosity scaling relations. We detect 447 clusters and determine 246 r2500 temperatures across all redshifts. Within 0.1 < z < 0.35 we find that r2500 Tx scales with optical richness as ln(kB Tx / 1.0 keV) = (0.52 \pm 0.05) ln(λ/70) + (1.85 \pm 0.03) with intrinsic scatter of 0.27 \pm 0.02 (1 σ). We investigate the distribution of offsets between the X-ray center and redMaPPer center within 0.1 < z < 0.35, finding that 68.3 \pm 6.5% of clusters are well-centered. However, we find a broad tail of large offsets in this distribution, and we explore some of the causes of redMaPPer miscentering.

astro-ph.CO↗

arXiv2019

Detection of anti-correlation of hot and cold baryons in galaxy clusters

The largest clusters of galaxies in the Universe contain vast amounts of dark matter, plus baryonic matter in two principal phases, a majority hot gas component and a minority cold stellar phase comprising stars, compact objects, and low-temperature gas. Hydrodynamic simulations indicate that the highest-mass systems retain the cosmic fraction of baryons, a natural consequence of which is anti-correlation between the masses of hot gas and stars within dark matter halos of fixed total mass. We report observational detection of this anti-correlation based on 4 elements of a $9\times9$ element covariance matrix for nine cluster properties, measured from X-ray, optical, infrared and millimetre wavelength observations of 41 clusters from the Local Cluster Substructure Survey. These clusters were selected using explicit and quantitative selection rules that were then encoded in our hierarchical Bayesian model. Our detection of anti-correlation is consistent with predictions from contemporary hydrodynamic cosmological simulations that were not tuned to reproduce this signal.

astro-ph.CO↗

arXiv2019

Percent-Level Test of Isotropic Expansion Using Type Ia Supernovae

We propose and implement a novel, robust, and non-parametric test of statistical isotropy of the expansion of the universe, and apply it to around one thousand type Ia supernovae from the Pantheon sample. We calculate the angular clustering of supernova magnitude residuals and compare it to the noise expected under the isotropic assumption. We also test for systematic effects and demonstrate that their effects are negligible or are already accounted for in our procedure. We express our constraints as an upper limit on the rms spatial variation in the Hubble parameter at late times. For the sky smoothed with a Gaussian with fwhm=60 deg, less than 1% rms spatial variation in the Hubble parameter is allowed at 99.7% confidence.

astro-ph.CO↗

arXiv2019

Dark Matter Science in the Era of LSST

Astrophysical observations currently provide the only robust, empirical measurements of dark matter. In the coming decade, astrophysical observations will guide other experimental efforts, while simultaneously probing unique regions of dark matter parameter space. This white paper summarizes astrophysical observations that can constrain the fundamental physics of dark matter in the era of LSST. We describe how astrophysical observations will inform our understanding of the fundamental properties of dark matter, such as particle mass, self-interaction strength, non-gravitational interactions with the Standard Model, and compact object abundances. Additionally, we highlight theoretical work and experimental/observational facilities that will complement LSST to strengthen our understanding of the fundamental characteristics of dark matter.

astro-ph.CO↗

arXiv2019

LoCuSS: Scaling relations between galaxy cluster mass, gas, and stellar content

We present a simultaneous analysis of galaxy cluster scaling relations between weak-lensing mass and multiple cluster observables, across a wide range of wavelengths, that probe both gas and stellar content. Our new hierarchical Bayesian model simultaneously considers the selection variable alongside all other observables in order to explicitly model intrinsic property covariance and account for selection effects. We apply this method to a sample of 41 clusters at $0.15<z<0.30$, with a well-defined selection criteria based on RASS X-ray luminosity, and observations from Chandra / XMM, SZA, Planck, UKIRT, SDSS and Subaru. These clusters have well-constrained weak-lensing mass measurements based on Subaru / Suprime-Cam observations, which serve as the reference masses in our model. We present 30 scaling relation parameters for 10 properties. All relations probing the intracluster gas are slightly shallower than self-similar predictions, in moderate tension with prior measurements, and the stellar fraction decreases with mass. K-band luminosity has the lowest intrinsic scatter with a 95th percentile of 0.16, while the lowest scatter gas probe is gas mass with a fractional intrinsic scatter of $0.16 \pm 0.03$. We find no distinction between the core-excised X-ray or high-resolution Sunyaev-Zel'dovich relations of clusters of different central entropy, but find with modest significance that higher entropy clusters have higher stellar fractions than their lower entropy counterparts. We also report posterior mass estimates from our likelihood model.

astro-ph.CO↗

arXiv2018

ActiveRemediation: The Search for Lead Pipes in Flint, Michigan

We detail our ongoing work in Flint, Michigan to detect pipes made of lead and other hazardous metals. After elevated levels of lead were detected in residents' drinking water, followed by an increase in blood lead levels in area children, the state and federal governments directed over $125 million to replace water service lines, the pipes connecting each home to the water system. In the absence of accurate records, and with the high cost of determining buried pipe materials, we put forth a number of predictive and procedural tools to aid in the search and removal of lead infrastructure. Alongside these statistical and machine learning approaches, we describe our interactions with government officials in recommending homes for both inspection and replacement, with a focus on the statistical model that adapts to incoming information. Finally, in light of discussions about increased spending on infrastructure development by the federal government, we explore how our approach generalizes beyond Flint to other municipalities nationwide.

cs.LG↗

arXiv2017

Localized massive halo properties in Bahamas and Macsis simulations: scalings, log-normality, and covariance

Using tens of thousands of halos realized in the BAHAMAS and MACSIS simulations produced with a consistent astrophysics treatment that includes AGN feedback, we validate a multi-property statistical model for the stellar and hot gas mass behavior in halos hosting groups and clusters of galaxies. The large sample size allows us to extract fine-scale mass--property relations (MPRs) by performing local linear regression (LLR) on individual halo stellar mass (${\rm M}_{\rm star}$) and hot gas mass (${\rm M}_{\rm gas}$) as a function of total halo mass (${\rm M}_{\rm halo}$). We find that: 1) both the local slope and variance of the MPRs run with mass (primarily) and redshift (secondarily); 2) the conditional likelihood, $p({\rm M}_{\rm star},\ {\rm M}_{\rm gas} | \ {\rm M}_{\rm halo}, z)$ is accurately described by a multivariate, log-normal distribution, and; 3) the covariance of ${\rm M}_{\rm star}$ and ${\rm M}_{\rm gas}$ at fixed ${\rm M}_{\rm halo}$ is generally negative, reflecting a partially closed baryon box model for high mass halos. We validate the analytical population model of Evrard et al. (2014), finding sub-percent accuracy in the log-mean halo mass selected at fixed property, $\langle \ln {\rm M}_{\rm halo} | {\rm M}_{\rm gas} \rangle$ or $\langle \ln {\rm M}_{\rm halo} | {\rm M}_{\rm star} \rangle$, when scale-dependent MPR parameters are employed. This work highlights the potential importance of allowing for running in the slope and scatter of MPRs when modeling cluster counts for cosmological studies. We tabulate LLR fit parameters as a function of halo mass at $z=0$, $0.5$ and 1 for two popular mass conventions.

astro-ph.CO↗

arXiv2017

Driving with Data: Modeling and Forecasting Vehicle Fleet Maintenance in Detroit

The City of Detroit maintains an active fleet of over 2500 vehicles, spending an annual average of over \$5 million on new vehicle purchases and over \$7.7 million on maintaining this fleet. Understanding the existence of patterns and trends in this data could be useful to a variety of stakeholders, particularly as Detroit emerges from Chapter 9 bankruptcy, but the patterns in such data are often complex and multivariate and the city lacks dedicated resources for detailed analysis of this data. This work, a data collaboration between the Michigan Data Science Team (http://midas.umich.edu/mdst) and the City of Detroit's Operations and Infrastructure Group, seeks to address this unmet need by analyzing data from the City of Detroit's entire vehicle fleet from 2010-2017. We utilize tensor decomposition techniques to discover and visualize unique temporal patterns in vehicle maintenance; apply differential sequence mining to demonstrate the existence of common and statistically unique maintenance sequences by vehicle make and model; and, after showing these time-dependencies in the dataset, demonstrate an application of a predictive Long Short Term Memory (LSTM) neural network model to predict maintenance sequences. Our analysis shows both the complexities of municipal vehicle fleet data and useful techniques for mining and modeling such data.

cs.CY↗

arXiv2017

A Data Science Approach to Understanding Residential Water Contamination in Flint

When the residents of Flint learned that lead had contaminated their water system, the local government made water-testing kits available to them free of charge. The city government published the results of these tests, creating a valuable dataset that is key to understanding the causes and extent of the lead contamination event in Flint. This is the nation's largest dataset on lead in a municipal water system. In this paper, we predict the lead contamination for each household's water supply, and we study several related aspects of Flint's water troubles, many of which generalize well beyond this one city. For example, we show that elevated lead risks can be (weakly) predicted from observable home attributes. Then we explore the factors associated with elevated lead. These risk assessments were developed in part via a crowd sourced prediction challenge at the University of Michigan. To inform Flint residents of these assessments, they have been incorporated into a web and mobile application funded by \texttt{Google.org}. We also explore questions of self-selection in the residential testing program, examining which factors are linked to when and how frequently residents voluntarily sample their water.

cs.LG↗