arXiv subjects

Tamás Szklenár

Publications and source records attributed to Tamás Szklenár.

5 recordsLinked to original sources

arXiv2025

Multiband neural network classification of ZTF light curves as LSST proxies

In this project we use data obtained by Zwicky Transient Facility to develop and test a neural-network-based, multiband classification algorithm to classify periodic variable stars (i.e. pulsating variable stars and eclipsing binaries). The aim is to utilize the algorithm on LSST data once they become available. Phase-folded light curve images and period information were used from five different variable star types: Classical and Type II Cepheids, δ Scuti stars, eclipsing binaries, and RR Lyrae stars. The data is taken from the 17th data release of ZTF, from which we used two passbands, g and r in this project. The periods were calculated from the raw data and this information was used as an additional numerical input in the neural network. For the training and testing process a supervised machine learning method was created, the neural network contains Convolutional Neural Networks concatenated with Fully Connected Layers. During the training-validation process the training accuracy reached 99% and the validation accuracy peaked at 95.6%. At the test classification phase three variable star types out of the 5 classes were classified with around 99% of accuracy, the other two also had very high accuracy, 89.6% and 93.6%.

astro-ph.IM↗

arXiv2025

Detection of Oscillation-like Patterns in Eclipsing Binary Light Curves using Neural Network-based Object Detection Algorithms

The primary aim of this research is to evaluate several convolutional neural network-based object detection algorithms for identifying oscillation-like patterns in light curves of eclipsing binaries. This involves creating a robust detection framework that can effectively process both synthetic light curves and real observational data. The study employs several state-of-the-art object detection algorithms, including Single Shot MultiBox Detector, Faster Region-based Convolutional Neural Network, You Only Look Once, and EfficientDet besides a custom non-pretrained model implemented from scratch. Synthetic light curve images and images derived from observational TESS light curves of known eclipsing binaries with a pulsating component were constructed with corresponding annotation files using custom scripts. The models were trained and validated on established datasets, followed by testing on unseen {\it{Kepler}} data to assess their generalization performance. The statistical metrics are also calculated to review the quality of each model. The results indicate that the pre-trained models exhibit high accuracy and reliability in detecting the targeted patterns. Faster R-CNN and You Only Look Once, in particular, showed superior performance in terms of object detection evaluation metrics on the validation dataset such as mAP value exceeding 99\%. Single Shot MultiBox Detector, on the other hand, is the fastest although it shows slightly lower performance with a mAP of 97\%. These findings highlight the potential of these models to contribute significantly to the automated determination of pulsating components in eclipsing binary systems, facilitating more efficient and comprehensive astrophysical investigations.

astro-ph.SR↗

arXiv2024

A theoretical framework for BL Her stars III. A case study: Robust light curve optimisation in the LMC

We carry out an extensive light curve comparison of BL Her stars using observations from Gaia DR3 and stellar pulsation models computed using MESA-RSP with the goal to obtain the best-matched modeled-observed pairs for BL Her stars in the LMC. We use the Fourier decomposition technique to analyse the light curves in the G band obtained from Gaia DR3 and from MESA-RSP and use a robust light curve fitting approach to score the modeled-observed pairs with respect to their pulsation periods and over their Fourier parameter space. We obtain the best-fit models for 48 BL Her stars in the LMC and thereby provide the stellar parameter estimates of these stars, 30 of which are labelled as the gold sample with superior light curve fits. We find a relatively flat distribution of stellar masses between 0.5-0.65 Msolar for the gold sample of modeled-observed pairs. An interesting result is that the majority of the best-matched models in the gold sample are computed using the convection parameter sets without radiative cooling. The period-Wesenheit relation for the best-matched gold sample of 30 BL Her models exhibits a slope of $-2.805 \pm 0.164$ while the corresponding period-radius relation exhibits a slope of $0.565 \pm 0.035$, both in good agreement with the empirical PW and PR slopes from BL Her stars in the LMC, respectively. We also used the Wesenheit magnitudes of the 30 best-matched modeled-observed pairs to estimate a distance modulus of $μ_{\rm LMC} = 18.582 \pm 0.067$ to the LMC, which lies within the bounds of previous literature values. We also discuss the degeneracy in the stellar parameters of the BL Her models that result in similar pulsation periods and light curve structure, and highlight that caution must be exercised while using the stellar parameter estimates.

astro-ph.SR↗

arXiv2023

Variable stars in the residual light curves of OGLE-IV eclipsing binaries towards the Galactic Bulge

Context. The Optical Gravitational Lensing Experiment (OGLE) observed around 450,000 eclipsing binaries (EBs) towards the Galactic Bulge. Decade-long photometric observations such as these provide an exceptional opportunity to thoroughly examine the targets. However, observing dense stellar fields such as the Bulge may result in blends and contamination by close objects. Aims. We searched for periodic variations in the residual light curves of EBs in OGLE-IV and created a new catalogue for the EBs that contain `background' signals after the investigation of the source of the signal. Methods. From the about half a million EB systems, we selected those that contain more than 4000 data points. We fitted the EB signal with a simple model and subtracted it. To identify periodical signals in the residuals, we used a GPU-based phase dispersion minimisation python algorithm called cuvarbase and validated the found periods with Lomb-Scargle periodograms. We tested the reliability of our method with artificial light curves. Results. We identified 354 systems where short-period background variation was significant. In these cases, we determined whether it is a new variable or just the result of contamination by an already catalogued nearby one. We classified 292 newly found variables into EB, $δ$ Scuti, or RR Lyrae categories, or their sub-classes, and collected them in a catalogue. We also discovered four new doubly eclipsing systems and one eclipsing multiple system with a $δ$ Scuti variable, and modelled the outer orbits of the components.

astro-ph.SR↗

arXiv2021

Finding flares in Kepler and TESS data with recurrent deep neural networks

Stellar flares are an important aspect of magnetic activity -- both for stellar evolution and circumstellar habitability viewpoints - but automatically and accurately finding them is still a challenge to researchers in the Big Data era of astronomy. We present an experiment to detect flares in space-borne photometric data using deep neural networks. Using a set of artificial data and real photometric data we trained a set of neural networks, and found that the best performing architectures were the recurrent neural networks (RNNs) using Long Short-Term Memory (LSTM) layers. The best trained network detected flares over {$5σ$ with $\gtrsim80$\% recall and precision} and was also capable to distinguish typical false signals (e.g. maxima of RR Lyr stars) from real flares. Testing the network trained on Kepler data on TESS light curves showed that the neural net is able to generalize and find flares - with similar effectiveness - in completely new data having previously unseen sampling and characteristics.

astro-ph.SR↗