arXiv subjects

P. Protopapas

Publications and source records attributed to P. Protopapas.

At least 19 recordsLinked to original sources

arXiv2026

Physics-Informed Neural Networks to Infer the Perpendicular Energy Conductivity in the Scrape-Off Layer of Stellarator Devices

In this work, we develop an inverse Physics-Informed Neural Network (PINN) framework to infer the dependence of the scrape-off layer (SOL) perpendicular heat conductivity on plasma density and temperature, $\kappa_\perp(n,T)$. The method combines radial profile measurements of electron density and temperature with the residual of a reduced one-dimensional SOL transport equation, so that the inferred conductivity is constrained by both the measurements and the underlying transport model. Three neural networks are trained simultaneously: two reconstruct the temperature and density profiles as functions of the radial coordinate and transported power, while a third represents the effective conductivity as a function of the local density and temperature. The framework is first validated using synthetic data generated from a prescribed conductivity function, allowing the inferred $\kappa_\perp(n,T)$ to be compared directly with the ground truth. The model recovers the imposed functional dependence with errors below $10~\%$ in the data-constrained region. Bootstrap resampling is shown to provide a practical indicator of prediction reliability and consistency. A scan in the number of plasma profiles used for training and the number of radial measurement positions per profile identifies a practical trade-off between reconstruction accuracy and data availability. Finally, the method is applied to an experimental dataset from the TJ-II stellarator obtained with the helium-beam diagnostic. This exploratory application provides an initial estimate of the effective SOL conductivity and illustrates the potential of inverse PINNs for extracting transport information from plasma edge measurements.

physics.plasm-ph↗

arXiv2026

On the performance of pre-trained vision transformers for supernova spectral classification using different spectral representations

The spectroscopic classification of supernovae is a key component of time-domain astronomy and plays an important role in the identification of Type Ia events. The increasing volume and diversity of spectroscopic data motivate the development of automated classification approaches. In this work, we explore the use of Transformer-based vision models for supernova spectral classification, focusing on how different visual representations of spectra and fine-tuning strategies affect classification performance when image-based architectures are applied to intrinsically one-dimensional data. We consider a dataset of 4,011 supernova spectra, augmented and rebalanced into three astrophysically motivated classes: normal Type Ia, other Type Ia subtypes, and core-collapse supernovae. Spectra are encoded as line-plot images using direct flux-wavelength visualizations as well as alternative flux-difference versus wavelength-difference heatmap representations designed to emphasize differential spectral structure. We evaluate several pre-trained Transformer architectures, including Vision Transformers (ViT), Swin Transformers, and a DINOv3-pretrained ViT, and examine the effects of fine-tuning depth, visual representation, and hyperparameter choices. We find that a plain flux-wavelength line plot outperforms the purpose-built maps on average, though log-scale maps remain competitive for specific architecture and fine-tuning combinations. On the held-out test set, our best model (\texttt{vitb-p16}) reaches a macro-F1 score of 86.3%, with per-class F1 scores of 94% for normal Ia, 77% for other Ia subtypes, and 88% for core-collapse supernovae. The dominant misclassification arises from confusion between Ia-91T and normal Ia spectra. These results demonstrate that Transformer-based vision models can provide competitive performance for single-spectrum supernova spectral classification.

astro-ph.IM↗

arXiv2025

Multiband Embeddings of Light Curves

In this work, we propose a novel ensemble of recurrent neural networks (RNNs) that considers the multiband and non-uniform cadence without having to compute complex features. Our proposed model consists of an ensemble of RNNs, which do not require the entire light curve to perform inference, making the inference process simpler. The ensemble is able to adapt to varying numbers of bands, tested on three real light curve datasets, namely Gaia, Pan-STARRS1, and ZTF, to demonstrate its potential for generalization. We also show the capabilities of deep learning to perform not only classification, but also regression of physical parameters such as effective temperature and radius. Our ensemble model demonstrates superior performance in scenarios with fewer observations, thus providing potential for early classification of sources from facilities such as Vera C. Rubin Observatory's LSST. The results underline the model's effectiveness and flexibility, making it a promising tool for future astronomical surveys. Our research has shown that a multitask learning approach can enrich the embeddings obtained by the models, making them instrumental to solve additional tasks, such as determining the orbital parameters of binary systems or estimating parameters for object types beyond periodic ones.

astro-ph.IM↗

arXiv2022

Semi-Supervised Classification and Clustering Analysis for Variable Stars

The immense amount of time series data produced by astronomical surveys has called for the use of machine learning algorithms to discover and classify several million celestial sources. In the case of variable stars, supervised learning approaches have become commonplace. However, this needs a considerable collection of expert-labeled light curves to achieve adequate performance, which is costly to construct. To solve this problem, we introduce two approaches. First, a semi-supervised hierarchical method, which requires substantially less trained data than supervised methods. Second, a clustering analysis procedure that finds groups that may correspond to classes or sub-classes of variable stars. Both methods are primarily supported by dimensionality reduction of the data for visualization and to avoid the curse of dimensionality. We tested our methods with catalogs collected from OGLE, CSS, and Gaia surveys. The semi-supervised method reaches a performance of around 90\% for all of our three selected catalogs of variable stars using only $5\%$ of the data in the training. This method is suitable for classifying the main classes of variable stars when there is only a small amount of training data. Our clustering analysis confirms that most of the clusters found have a purity over 90\% with respect to classes and 80\% with respect to sub-classes, suggesting that this type of analysis can be used in large-scale variability surveys as an initial step to identify which classes or sub-classes of variable stars are present in the data and/or to build training sets, among many other possible applications.

astro-ph.SR↗

arXiv2022

ASTROMER: A transformer-based embedding for the representation of light curves

Taking inspiration from natural language embeddings, we present ASTROMER, a transformer-based model to create representations of light curves. ASTROMER was pre-trained in a self-supervised manner, requiring no human-labeled data. We used millions of R-band light sequences to adjust the ASTROMER weights. The learned representation can be easily adapted to other surveys by re-training ASTROMER on new sources. The power of ASTROMER consists of using the representation to extract light curve embeddings that can enhance the training of other models, such as classifiers or regressors. As an example, we used ASTROMER embeddings to train two neural-based classifiers that use labeled variable stars from MACHO, OGLE-III, and ATLAS. In all experiments, ASTROMER-based classifiers outperformed a baseline recurrent neural network trained on light curves directly when limited labeled data was available. Furthermore, using ASTROMER embeddings decreases computational resources needed while achieving state-of-the-art results. Finally, we provide a Python library that includes all the functionalities employed in this work. The library, main code, and pre-trained weights are available at https://github.com/astromer-science

astro-ph.IM↗

arXiv2021

The effect of phased recurrent units in the classification of multiple catalogs of astronomical lightcurves

In the new era of very large telescopes, where data is crucial to expand scientific knowledge, we have witnessed many deep learning applications for the automatic classification of lightcurves. Recurrent neural networks (RNNs) are one of the models used for these applications, and the LSTM unit stands out for being an excellent choice for the representation of long time series. In general, RNNs assume observations at discrete times, which may not suit the irregular sampling of lightcurves. A traditional technique to address irregular sequences consists of adding the sampling time to the network's input, but this is not guaranteed to capture sampling irregularities during training. Alternatively, the Phased LSTM unit has been created to address this problem by updating its state using the sampling times explicitly. In this work, we study the effectiveness of the LSTM and Phased LSTM based architectures for the classification of astronomical lightcurves. We use seven catalogs containing periodic and nonperiodic astronomical objects. Our findings show that LSTM outperformed PLSTM on 6/7 datasets. However, the combination of both units enhances the results in all datasets.

astro-ph.IM↗

arXiv2020

The Automatic Learning for the Rapid Classification of Events (ALeRCE) Alert Broker

We introduce the Automatic Learning for the Rapid Classification of Events (ALeRCE) broker, an astronomical alert broker designed to provide a rapid and self--consistent classification of large etendue telescope alert streams, such as that provided by the Zwicky Transient Facility (ZTF) and, in the future, the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). ALeRCE is a Chilean--led broker run by an interdisciplinary team of astronomers and engineers, working to become intermediaries between survey and follow--up facilities. ALeRCE uses a pipeline which includes the real--time ingestion, aggregation, cross--matching, machine learning (ML) classification, and visualization of the ZTF alert stream. We use two classifiers: a stamp--based classifier, designed for rapid classification, and a light--curve--based classifier, which uses the multi--band flux evolution to achieve a more refined classification. We describe in detail our pipeline, data products, tools and services, which are made public for the community (see \url{https://alerce.science}). Since we began operating our real--time ML classification of the ZTF alert stream in early 2019, we have grown a large community of active users around the globe. We describe our results to date, including the real--time processing of $9.7\times10^7$ alerts, the stamp classification of $1.9\times10^7$ objects, the light curve classification of $8.5\times10^5$ objects, the report of 3088 supernova candidates, and different experiments using LSST-like alert streams. Finally, we discuss the challenges ahead to go from a single-stream of alerts such as ZTF to a multi--stream ecosystem dominated by LSST.

astro-ph.IM↗

arXiv2019

Physical Symmetries Embedded in Neural Networks

Neural networks are a central technique in machine learning. Recent years have seen a wave of interest in applying neural networks to physical systems for which the governing dynamics are known and expressed through differential equations. Two fundamental challenges facing the development of neural networks in physics applications is their lack of interpretability and their physics-agnostic design. The focus of the present work is to embed physical constraints into the structure of the neural network to address the second fundamental challenge. By constraining tunable parameters (such as weights and biases) and adding special layers to the network, the desired constraints are guaranteed to be satisfied without the need for explicit regularization terms. This is demonstrated on upervised and unsupervised networks for two basic symmetries: even/odd symmetry of a function and energy conservation. In the supervised case, the network with embedded constraints is shown to perform well on regression problems while simultaneously obeying the desired constraints whereas a traditional network fits the data but violates the underlying constraints. Finally, a new unsupervised neural network is proposed that guarantees energy conservation through an embedded symplectic structure. The symplectic neural network is used to solve a system of energy-conserving differential equations and out-performs an unsupervised, non-symplectic neural network.

physics.comp-ph↗

arXiv2014

Pan-STARRS 1 observations of the unusual active Centaur P/2011 S1(Gibbs)

P/2011 S1 (Gibbs) is an outer solar system comet or active Centaur with a similar orbit to that of the famous 29P/Schwassmann-Wachmann 1. P/2011 S1 (Gibbs) has been observed by the Pan-STARRS 1 (PS1) sky survey from 2010 to 2012. The resulting data allow us to perform multi-color studies of the nucleus and coma of the comet. Analysis of PS1 images reveals that P/2011 S1 (Gibbs) has a small nucleus $< 4$ km radius, with colors $g_{P1}-r_{P1} = 0.5 \pm 0.02$, $r_{P1}-i_{P1} = 0.12 \pm 0.02$ and $i_{P1}-z_{P1} = 0.46 \pm 0.03$. The comet remained active from 2010 to 2012, with a model-dependent mass-loss rate of $\sim100$ kg s$^{-1}$. The mass-loss rate per unit surface area of P/2011 S1 (Gibbs) is as high as that of 29P/Schwassmann-Wachmann 1, making it one of the most active Centaurs. The mass-loss rate also varies with time from $\sim 40$ kg s$^{-1}$ to 150 kg s$^{-1}$. Due to its rather circular orbit, we propose that P/2011 S1 (Gibbs) has 29P/Schwassmann-Wachmann 1-like outbursts that control the outgassing rate. The results indicate that it may have a similar surface composition to that of 29P/Schwassmann-Wachmann 1. Our numerical simulations show that the future orbital evolution of P/2011 S1 (Gibbs) is more similar to that of the main population of Centaurs than to that of 29P/Schwassmann-Wachmann 1. The results also demonstrate that P/2011 S1 (Gibbs) is dynamically unstable and can only remain near its current orbit for roughly a thousand years.

astro-ph.EP↗

arXiv2013

The TAOS Project: Results From Seven Years of Survey Data

The Taiwanese-American Occultation Survey (TAOS) aims to detect serendipitous occultations of stars by small (about 1 km diameter) objects in the Kuiper Belt and beyond. Such events are very rare (<0.001 events per star per year) and short in duration (about 200 ms), so many stars must be monitored at a high readout cadence. TAOS monitors typically around 500 stars simultaneously at a 5 Hz readout cadence with four telescopes located at Lulin Observatory in central Taiwan. In this paper, we report the results of the search for small Kuiper Belt Objects (KBOs) in seven years of data. No occultation events were found, resulting in a 95% c.l. upper limit on the slope of the faint end of the KBO size distribution of q = 3.34 to 3.82, depending on the surface density at the break in the size distribution at a diameter of about 90 km.

astro-ph.EP↗

arXiv2010

The TAOS Project Stellar Variability II. Detection of 15 Variable Stars

The Taiwanese-American Occultation Survey (TAOS) project has collected more than a billion photometric measurements since 2005 January. These sky survey data-covering timescales from a fraction of a second to a few hundred days-are a useful source to study stellar variability. A total of 167 star fields, mostly along the ecliptic plane, have been selected for photometric monitoring with the TAOS telescopes. This paper presents our initial analysis of a search for periodic variable stars from the time-series TAOS data on one particular TAOS field, No. 151 (RA = 17$^{\rm h}30^{\rm m}6\fs$67, Dec = 27\degr17\arcmin 30\arcsec, J2000), which had been observed over 47 epochs in 2005. A total of 81 candidate variables are identified in the 3 square degree field, with magnitudes in the range 8 < R < 16. On the basis of the periodicity and shape of the lightcurves, 29 variables, 15 of which were previously unknown, are classified as RR Lyrae, Cepheid, delta Scuti, SX Phonencis, semi-regular and eclipsing binaries.

astro-ph.SR↗

arXiv2010

The TAOS Project: Statistical Analysis of Multi-Telescope Time Series Data

The Taiwanese-American Occultation Survey (TAOS) monitors fields of up to ~1000 stars at 5 Hz simultaneously with four small telescopes to detect occultation events from small (~1 km) Kuiper Belt Objects (KBOs). The survey presents a number of challenges, in particular the fact that the occultation events we are searching for are extremely rare and are typically manifested as slight flux drops for only one or two consecutive time series measurements. We have developed a statistical analysis technique to search the multi-telescope data set for simultaneous flux drops which provides a robust false positive rejection and calculation of event significance. In this paper, we describe in detail this statistical technique and its application to the TAOS data set.

astro-ph.EP↗

arXiv2010

The TAOS Project: Upper Bounds on the Population of Small KBOs and Tests of Models of Formation and Evolution of the Outer Solar System

We have analyzed the first 3.75 years of data from TAOS, the Taiwanese American Occultation Survey. TAOS monitors bright stars to search for occultations by Kuiper Belt Objects (KBOs). This dataset comprises 5e5 star-hours of multi-telescope photometric data taken at 4 or 5 Hz. No events consistent with KBO occultations were found in this dataset. We compute the number of events expected for the Kuiper Belt formation and evolution models of Pan & Sari (2005), Kenyon & Bromley (2004), Benavidez & Campo Bagatin (2009), and Fraser (2009). A comparison with the upper limits we derive from our data constrains the parameter space of these models. This is the first detailed comparison of models of the KBO size distribution with data from an occultation survey. Our results suggest that the KBO population is comprised of objects with low internal strength and that planetary migration played a role in the shaping of the size distribution.

astro-ph.EP↗

arXiv2009

The TAOS Project Stellar Variability I. Detection of Low-Amplitude delta Scuti Stars

We analyzed data accumulated during 2005 and 2006 by the Taiwan-American Occultation Survey (TAOS) in order to detect short-period variable stars (periods of <~ 1 hour) such as delta Scuti. TAOS is designed for the detection of stellar occultation by small-size Kuiper Belt Objects (KBOs) and is operating four 50cm telescopes at an effective cadence of 5Hz. The four telescopes simultaneously monitor the same patch of the sky in order to reduce false positives. To detect short-period variables, we used the Fast Fourier Transform algorithm (FFT) inasmuch as the data points in TAOS light-curves are evenly spaced. Using FFT, we found 41 short-period variables with amplitudes smaller than a few hundredths of a magnitude and periods of about an hour, which suggest that they are low-amplitude delta Scuti stars (LADS). The light-curves of TAOS delta Scuti stars are accessible online at the Time Series Center website (http://timemachine.iic.harvard.edu)

astro-ph.SR↗

arXiv2009

Searching for sub-kilometer TNOs using Pan-STARRS video mode lightcurves: Preliminary study and evaluation using engineering data

We present a pre-survey study of using Pan-STARRS high sampling rate video mode guide star images to search for TNOs. With suitable selection of the guide stars within the Pan-STARRS 7 deg^{2} field of view, the lightcurves of these guide stars can also be used to search for occultations by TNOs. The best target stars for this purpose are stars with high signal-to-noise ratio (SNR) and small angular size.In order to do this, we compiled a catalog using the SNR calculated from stars with m_V <13 mag in the Tycho2 catalog then cross matched these stars with the 2MASS catalog and estimated their angular sizes from (V-K) color. We also outlined a new detection method based on matched filter that is optimized to search for diffraction patterns in the lightcurves due to occultation by sub-kilometer TNOs. A detection threshold is set to compromise between real detections and false positives. Depending on the theoretical size distribution model used, we expect to find up to a hundred events during the three-year life time of the Pan-STARRS-1 project. We have tested the detection algorithm and the pipeline on a set of engineering data (taken at 10Hz in stead of 30Hz). No events were found within the engineering data, which is consistent with the small size of the data set and the theoretical models. Meanwhile, with a total of ~ 22 star-hours video mode data (|β| < 10^{\circ}), we are able to set an upper limit of N(>0.5 km) ~ 2.47x10^10 deg^-2 at 95% confidence limit.

astro-ph.EP↗

arXiv2009

Upper Limits on the Number of Small Bodies in Sedna-Like Orbits by the TAOS Project

We present the results of a search for occultation events by objects at distances between 100 and 1000 AU in lightcurves from the Taiwanese-American Occultation Survey (TAOS). We searched for consecutive, shallow flux reductions in the stellar lightcurves obtained by our survey between 7 February 2005 and 31 December 2006 with a total of $\sim4.5\times10^{9}$ three-telescope simultaneous photometric measurements. No events were detected, allowing us to set upper limits on the number density as a function of size and distance of objects in Sedna-like orbits, using simple models.

astro-ph.EP↗

arXiv2008

First Results From The Taiwanese-American Occultation Survey (TAOS)

Results from the first two years of data from the Taiwanese-American Occultation Survey (TAOS) are presented. Stars have been monitored photometrically at 4 Hz or 5 Hz to search for occultations by small (~3 km) Kuiper Belt Objects (KBOs). No statistically significant events were found, allowing us to present an upper bound to the size distribution of KBOs with diameters 0.5 km < D < 28 km.

astro-ph↗

arXiv2005

Combined reconstruction of weak and strong lensing data with WSLAP

We describe a method to estimate the mass distribution of a gravitational lens and the position of the sources from combined strong and weak lensing data. The algorithm combines weak and strong lensing data in a unified way producing a solution which is valid in both the weak and strong lensing regimes. We study how the result depends on the relative weighting of the weak and strong lensing data and on choice of basis to represent the mass distribution. We find that combining weak and strong lensing information has two major advantages: it eliminates the need for priors and/or regularization schemes for the intrinsic size of the background galaxies (this assumption was needed in previous strong lensing algorithms) and it corrects for biases in the recovered mass in the outer regions where the strong lensing data is less sensitive. The code is implemented into a software package called WSLAP (Weak & Strong Lensing Analysis Package) which is publicly available at http://darwin.cfa.harvard.edu/SLAP/

astro-ph↗