A5047358719- Galaxies: Formation, Evolution, Phenomena
- Astronomy and Astrophysical Research
- Model Reduction and Neural Networks
- Gaussian Processes and Bayesian Inference
- Computational Physics and Python Applications
- Cosmology and Gravitation Theories
- Advanced Image Processing Techniques
- Stellar, planetary, and galactic studies
- Fluid Dynamics and Turbulent Flows
- Advanced Vision and Imaging
- Adaptive optics and wavefront sensing
- Dark Matter and Cosmic Phenomena
- Scientific Research and Discoveries
- Advanced Multi-Objective Optimization Algorithms
- Astrophysics and Star Formation Studies
- Explainable Artificial Intelligence (XAI)
- Probabilistic and Robust Engineering Design
- Composite Material Mechanics
- Calibration and Measurement Techniques
- Domain Adaptation and Few-Shot Learning
- Machine Learning and Data Classification
- Astrophysics and Cosmic Phenomena
- Anomaly Detection Techniques and Applications
- Impact of Light on Environment and Health
- Reservoir Engineering and Simulation Methods
Alex's Lemonade Stand Foundation
2024
Argonne National Laboratory
2019-2024
University of Chicago
2020-2021
University of Kansas
2015-2017
The cosmic web is one of the most striking features distribution galaxies and dark matter on largest scales in Universe. It composed dense regions packed full galaxies, long filamentary bridges, flattened sheets vast low-density voids. study has focused primarily identification such features, understanding environmental effects galaxy formation halo assembly. As such, a variety different methods have been devised to classify – depending data at hand, be it numerical simulations, large sky...
We explore unsupervised machine learning for galaxy morphology analyses using a combination of feature extraction with vector-quantised variational autoencoder (VQ-VAE) and hierarchical clustering (HC). propose new methodology that includes: (1) consideration the performance simultaneously when features from images; (2) allowing various distance thresholds within HC algorithm; (3) orientation to determine number clusters. This setup provides 27 clusters created this which we show are well...
Testing a subset of viable cosmological models beyond General Relativity (GR), with implications for cosmic acceleration and the Dark Energy associated it, is within reach Rubin Observatory Legacy Survey Space Time (LSST) part its endeavor. Deviations from GR-w(z)CDM can manifest in growth rate structure lensing, as well screening effects on non-linear scales. We explore constraining power small-scale deviations predicted by f(R) Hu-Sawicki Modified Gravity (MG) candidate, emulating this...
We consider the use of probabilistic neural networks for fluid flow {surrogate modeling} and data recovery. This framework is constructed by assuming that target variables are sampled from a Gaussian distribution conditioned on inputs. Consequently, overall formulation sets up procedure to predict hyperparameters this which then used compute an objective function given training data. demonstrate has ability provide prediction confidence intervals based assumption posterior, appropriate model...
Carbon-enhanced metal-poor (CEMP) stars comprise almost a third of with [Fe/H] < --2, although their origins are still poorly understood. It is highly likely that one sub-class (CEMP-$s$ stars) tied to mass-transfer events in binary stars, while another (CEMP-no enriched by the nucleosynthetic yields first generations stars. Previous studies CEMP have primarily concentrated on Galactic halo, but more recently they also been detected thick disk and bulge components Milky Way. $Gaia$ DR3 has...
The problem of anomaly detection in astronomical surveys is becoming increasingly important as data sets grow size. We present the results an unsupervised method using a Wasserstein generative adversarial network (WGAN) on nearly one million optical galaxy images Hyper Suprime-Cam (HSC) survey. WGAN learns to generate realistic HSC-like galaxies that follow distribution set; anomalous are defined based poor reconstruction by generator and outlying features learned discriminator. find...
We report the results of first study multi-stream environment dark matter haloes in cosmological N-body simulations LCDM cosmology. The full dynamical state can be described as a three-dimensional sub-manifold six dimensional phase space - sheet. In our we use Lagrangian x = x(q,t) (where and q are co-moving Eulerian coordinates respectively), which is dynamically equivalent to sheet but more convenient for numerical analysis. Our major summarized follows. At resolution simulation i.e....
Optical astronomical images are strongly affected by the point spread function (PSF) of optical system and atmosphere (seeing) which blurs observed image. The amount blurring depends both on band, atmospheric conditions during observation. A typical image will likely have a unique PSF, that is non-circular different in bands. At same time, observations known stars also give us an accurate determination this PSF. Therefore, any serious candidate for production analysis must take PSF into...
In simulation-based models of the galaxy-halo connection, theoretical predictions for galaxy clustering and lensing are typically made based on Monte Carlo realizations a mock universe. this paper, we use Subhalo Abundance Matching (SHAM) as toy model to introduce an alternative stochastic population, demonstrating how make that both exact differentiable with respect parameters model. Conventional implementations SHAM iterative algorithms such Richardson-Lucy deconvolution; here JAX library...
This work presents an efficient end-to-end dimensionality reduction based surrogate modeling framework that maps high-dimensional image-like input to output. approach contains two modules: 1) module reduce the and output from a space reduced or low-dimensional space, 2) construct fast accurate input-output relationship in space. In module, both linear (principal component analysis (PCA)) nonlinear methods (kernel principal (KPCA), uniform manifold approximation projection (UMAP)) are used....
<title>Abstract</title> AstroSage-Llama-3.1-8B is a domain-specialized natural-language AI assistant tailored for research in astronomy, astrophysics, and cosmology. Trained on the complete collection of astronomy-related arXiv papers from 2007-2024 along with millions synthetically-generated question-answer pairs other astronomical literature, demonstrates remarkable proficiency wide range questions. scores 80.9% AstroMLab-1 benchmark, greatly outperforming all models—proprietary...
Topological connections in the single-streaming voids and multistreaming filaments walls reveal a cosmic web structure different from traditional mass density fields. A single void not only percolates multistream field all directions, but also occupies over 99 per cent of regions. Sub-grid analyses on scales smaller than simulation resolution tiny pockets that are isolated by membranes structure. For excursion sets, percolating is significantly thinner overdensity approach. Hessian...
Abstract Establishing fast and accurate structure-to-property relationships is an important component in the design discovery of advanced materials. Physics-based simulation models like finite element method (FEM) are often used to predict deformation, stress, strain fields as a function material microstructure structural systems. Such may be computationally expensive time intensive if underlying physics system complex. This limits their application solve inverse problems identify structures...
ABSTRACT Photometric redshift estimation algorithms are often based on representative data from observational campaigns. Data-driven methods of this type subject to a number potential deficiencies, such as sample bias and incompleteness. Motivated by these considerations, we propose using physically motivated synthetic spectral energy distributions in estimation. In addition, the would have span domain colour-redshift space concordant with that targeted surveys. With matched distribution...
Large pristine samples of red clump stars are highly sought after given that they standard candles and give precise distances even at large distances. However, it is difficult to cleanly select clumps because can have the same T$_{\mathrm{eff}}$ log $g$ as giant branch stars. Recently, was shown asteroseismic parameters, $\rm{\Delta}$P $\rm{\Delta\nu}$, which used accurately stars, be derived from spectra using change in surface carbon nitrogen ratio ([C/N]) caused by mixing during branch....
Upcoming large astronomical surveys are expected to capture an unprecedented number of strong gravitational lensing systems. Deep learning is emerging as a promising practical tool for the detection and quantification these galaxy-scale image distortions. The absence quantities representative data from current motivates development robust forward-modeling approach using synthetic images. Using mock sample lenses created upon state-of-the-art extragalactic catalogs, we train modular deep...
Machine learning has rapidly become a tool of choice for the astronomical community. It is being applied across wide range wavelengths and problems, from classification transients to neural network emulators cosmological simulations, shifting paradigms about how we generate report scientific results. At same time, this class method comes with its own set best practices, challenges, drawbacks, which, at present, are often reported on incompletely in astrophysical literature. With paper, aim...
Estimating uncertainty is at the core of performing scientific measurements in HEP: a measurement not useful without an estimate its uncertainty. The goal quantification (UQ) inextricably linked to question, "how do we physically and statistically interpret these uncertainties?" answer this question depends only on computational task aim undertake, but also methods use for that task. For artificial intelligence (AI) applications HEP, there are several areas where interpretable UQ essential,...
Using Subaru Hyper Suprime-Cam (HSC) year 1 data, we perform the first $k$-cut cosmic shear analysis constraining both $\Lambda$CDM and $f(R)$ Hu-Sawicki modified gravity. To generate theory vector, use matter power spectrum emulator trained on COLA (COmoving Lagrangian Acceleration) simulations. The method is used to significantly down-weight sensitivity small scale ($k > \ h {\rm Mpc }^{-1}$) modes in where less accurate, while simultaneously ensuring our results are robust baryonic...
We present an anomaly detection method using Wasserstein generative adversarial networks (WGANs) on optical galaxy images from the wide-field survey conducted with Hyper Suprime-Cam (HSC) Subaru Telescope in Hawai'i. The WGAN is trained entire sample, and learns to generate realistic HSC-like that follow distribution of training data. identify which are less well-represented generator's latent space, discriminator flags as realistic; these thus anomalous respect rest propose a new approach...