A5084374765- Gamma-ray bursts and supernovae
- Astrophysics and Cosmic Phenomena
- Astronomical Observations and Instrumentation
- Galaxies: Formation, Evolution, Phenomena
- Astronomy and Astrophysical Research
- Advanced Statistical Methods and Models
- Gaussian Processes and Bayesian Inference
- Anomaly Detection Techniques and Applications
- CCD and CMOS Imaging Sensors
- Machine Learning and Data Classification
- Domain Adaptation and Few-Shot Learning
- Computational Physics and Python Applications
- Adaptive optics and wavefront sensing
- Infrared Target Detection Methodologies
- Astrophysical Phenomena and Observations
- Spectroscopy and Chemometric Analyses
- Pulsars and Gravitational Waves Research
- Seismic Imaging and Inversion Techniques
- Impact of Light on Environment and Health
- Image Processing Techniques and Applications
- Machine Learning and Algorithms
- Particle Detector Development and Performance
- Data Visualization and Analytics
- Time Series Analysis and Forecasting
- Statistical Methods and Inference
Fermi National Accelerator Laboratory
2020-2025
Netherlands Institute for Radio Astronomy
2023-2024
Argonne National Laboratory
2022
University of Chicago
2022
University of Belgrade
2015-2020
Mathematical Institute of the Serbian Academy of Sciences and Arts
2019-2020
Serbian Academy of Sciences and Arts
2019-2020
In astronomy, neural networks are often trained on simulation data with the prospect of being used telescope observations. Unfortunately, training a model and then applying it to instrument leads substantial potentially even detrimental decrease in accuracy new target dataset. Simulated represent different domains, for an algorithm work both, domain-invariant learning is necessary. Here we employ domain adaptation techniques$-$ Maximum Mean Discrepancy (MMD) as additional transfer loss...
The large number of strong lenses discoverable in future astronomical surveys will likely enhance the value gravitational lensing as a cosmic probe dark energy and matter. However, leveraging increased statistical power such samples require further development automated lens modeling techniques. We show that deep learning simulation-based inference (SBI) methods produce informative reliable estimates parameter posteriors for systems ground-based surveys. present examination comparison two...
Modern neural networks (NNs) often do not generalize well in the presence of a "covariate shift"; that is, situations where training and test data distributions differ, but conditional distribution classification labels remains unchanged. In such cases, NN generalization can be reduced to problem learning more domain-invariant features. Domain adaptation (DA) methods include range techniques aimed at achieving this; however, these have struggled with need for extensive hyperparameter tuning,...
Abstract Artificial intelligence methods show great promise in increasing the quality and speed of work with large astronomical datasets, but high complexity these leads to extraction dataset-specific, non-robust features. Therefore, such do not generalize well across multiple datasets. We present a universal domain adaptation method, DeepAstroUDA , as an approach overcome this challenge. This algorithm performs semi-supervised (DA) can be applied datasets different data distributions class...
Abstract With increased adoption of supervised deep learning methods for work with cosmological survey data, the assessment data perturbation effects (that can naturally occur in processing and analysis pipelines) development that increase model robustness are increasingly important. In context morphological classification galaxies, we study perturbations imaging data. particular, examine consequences using neural networks when training on baseline testing perturbed We consider associated...
Methods based on machine learning have recently made substantial inroads in many corners of cosmology. Through this process, new computational tools, perspectives data collection, model development, analysis, and discovery, as well communities educational pathways emerged. Despite rapid progress, potential at the intersection cosmology remains untapped. In white paper, we summarize current ongoing developments relating to application within provide a set recommendations aimed maximizing...
With the advent of billion-galaxy surveys with complex data, need hour is to efficiently model galaxy spectral energy distributions (SEDs) robust uncertainty quantification. The combination Simulation-Based inference (SBI) and amortized Neural Posterior Estimation (NPE) has been successfully used analyse simulated real photometry both precisely efficiently. In this work, we utilise build on existing literature noisy spectra. Here, demonstrate a proof-of-concept study spectra that a) an...
In this paper we present the updated empirical radio surface-brightness-to-diameter ($\Sigma$--$D$) relation for Galactic supernova remnants (SNRs) calibrated using $110$ SNRs with reliable distances. We apply orthogonal fitting procedure and kernel density smoothing in $\Sigma-D$ plane compare results latest theoretical relations derived from simulations of evolution SNRs. argue that best agreement between simulated is achieved if mixed-morphology both, low brightness small diameter, are...
Abstract Development of the Rubin Observatory Legacy Survey Space and Time (LSST) includes a series Data Challenges (DCs) arranged by various LSST Scientific Collaborations that are taking place during project's preoperational phase. The AGN Science Collaboration Challenge (AGNSC-DC) is partial prototype expected data on active galactic nuclei (AGNs), aimed at validating machine learning approaches for selection characterization in large surveys like LSST. AGNSC-DC took 2021, focusing...
the first time, we demonstrate successful use of domain adaptation on two very different observational datasets (from SDSS and DECaLS). We show that our method is capable bridging gap between astronomical surveys, also performs well for anomaly detection clustering unknown data in unlabeled dataset. apply model to examples galaxy morphology classification tasks with detection: 1) classifying spiral elliptical galaxies merging (three classes including one class); 2) a more granular problem...
Deep learning models have been shown to outperform methods that rely on summary statistics, like the power spectrum, in extracting information from complex cosmological data sets. However, due differences subgrid physics implementation and numerical approximations across different simulation suites, trained one show a drop performance when tested another. Similarly, any of simulations would also likely experience applied observational data. Training two suites CAMELS hydrodynamic...
In astronomy, neural networks are often trained on simulated data with the prospect of being applied to real observations. Unfortunately, simply training a deep network images from one domain does not guarantee satisfactory performance new different domain. The ability share cross-domain knowledge is main advantage modern adaptation techniques. Here we demonstrate use two techniques - Maximum Mean Discrepancy (MMD) and adversarial Domain Adversarial Neural Networks (DANN) for classification...
We present detection of 64 H II regions, three superbubbles and two optical supernova remnant (SNR) candidates in the nearby irregular galaxy NGC 2366. The SNR were detected by applying [S II]/H? ratio criterion to observations made with 2-m RCC telescope at Rozhen National Astronomical Observatory Bulgaria. In this paper we report coordinates, diameters, H? II] fluxes for objects across fields view 2366 galaxy. Using archival XMM-Newton suggest possible X-ray counterparts candidates. Also,...
Inferring the values and uncertainties of cosmological parameters in a cosmology model is paramount importance for modern cosmic observations. In this paper, we use simulation-based inference (SBI) approach to estimate constraints from simplified galaxy cluster observation analysis. Using data generated Quijote simulation suite analytical models, train machine learning algorithm learn probability function between possible observables. The posterior distribution at given then obtained by...
We present the detection of 16 optical supernova remnant (SNR) candidates in nearby spiral galaxy IC342. The were detected by applying [SII]/H$\alpha$ ratio criterion on observations made with 2 m RCC telescope at Rozhen National Astronomical Observatory Bulgaria. In this paper, we report coordinates, diameters, H$\alpha$ and [SII] fluxes for SNRs two fields view IC342 galaxy. Also, estimate that contamination total flux from observed portion is 1.4%. This would represent fractional error...
and classification, their applicability to massive spatio-temporal data sets, such as remote sensing data, has been limited owing the high computational costs involved. In this paper we address scalability aspect of GP based time series change detection. Specifically, exploit special structure covariance matrix generated for analysis come up with methods that can efficiently estimate hyper-parameters associated well identify changes in while requiring a memory footprint which is linear size...
galaxy-galaxy lenses. We demonstrate the successful application of Neural Posterior Estimation (NPE) to automate inference a 12-parameter lens mass model for DES-like ground-based imaging data. compare our NPE constraints Bayesian Network (BNN) and find that it outperforms BNN, producing posterior distributions are most part both more accurate precise; in particular, several source-light parameters systematically biased BNN implementation.
Within the decade, many new ground and space-based observatories will become operational, generating massive amounts of data on short timescales. New surveys like Rubin Observatory's Legacy Survey Space Time (LSST) be capable observing objects with greater resolution than ever before, but processing analyzing these datasets optimally pose a significant challenge. In an effort to prepare for this, we explore how incorporating Deep Neural Networks can better support future data-intensive...
In the boundless expanse of cosmos lies a tapestry mysteries waiting to be unraveled. this talk, we delve into realm "Cosmic Algorithms," where marriage cutting-edge artificial intelligence (AI) and astrophysical inquiry paves way for unprecedented discoveries. One foremost challenges face in contemporary astrophysics is staggering size complexity astronomical datasets. I will discuss how AI provides transformative solution these challenges, enabling us efficiently sift through vast amounts...