arXiv subjects

M. Yu. Vasylenko

Publications and source records attributed to M. Yu. Vasylenko.

5 recordsLinked to original sources

arXiv2026

Visual inspection of potential exocomet transits identified through machine learning and statistical methods

In this work, we explore several ways to detect possible exocomet transits in the TESS (The Transiting Exoplanet Survey Satellite) light curves. The first one has been presented in our previous work, a machine learning approach based on the Random Forest algorithm. It was trained on asymmetric transit profiles calculated as a result of the modelling of a comet transit, and then applied to real star light curves from Sector 1 of TESS. This allowed us to detect 32 candidates with weak and non-periodic brightness dips that may correspond to comet-like events. The aim of this work is to analyse the events identified by the visual inspection to make sure that the features detected were not caused by instrumental effects. The second approach to detect possible exocomet transits, which is proposed, is an independent statistical method to test the results of the machine learning algorithm and to look for asymmetric minima directly in the light curves. This approach was applied to \b{eta} Pictoris light curves using TESS data from Sectors 5, 6, 32, and 33. The algorithm reproduced nearly all previously known events deeper than 0.03 % of the star flux, showing that it is efficient to detect shallow and irregular flux changes in the different sectors of the TESS data and at the different levels of noise. The combination of machine learning, visual inspection, and statistical analysis facilitates the identification of faint and short-lived asymmetric transits in photometric data. Although the number of confirmed exocomet transits is still small, the growing amount of observations points to their likely presence in many young planetary systems.

astro-ph.EP↗

arXiv2024

Hunting for exocomet transits in the TESS database using the Random Forest method

This study introduces an approach to detecting exocomet transits in the dataset of the Transiting Exoplanet Survey Satellite (TESS), specifically within its Sector 1. Given the limited number of exocomet transits detected in the observed light curves, creating a sufficient training sample for the machine learning method was challenging. We developed a unique training sample by encapsulating simulated asymmetric transit profiles into observed light curves, thereby creating realistic data for the model training. To analyze these light curves, we employed the TSFresh software, which was a tool for extracting key features that were then used to refine our Random Forest model training. Considering that cometary transits typically exhibit a small depth, less than 1% of the star's brightness, we chose to limit our sample to the CDPP parameter. Our study focused on two target samples: light curves with a CDPP of less than 40 ppm and light curves with a CDPP of up to 150 ppm. Each sample was accompanied by a corresponding training set. This methodology achieved an accuracy of approximately 96%, with both precision and recall rates exceeding 95% and a balanced F1-score of around 96%. This level of accuracy was effective in distinguishing between 'exocomet candidate' and 'non-candidate' classifications for light curves with a CDPP of less than 40 ppm, and our model identified 12 potential exocomet candidates. However, when applying machine learning to less accurate light curves (CDPP up to 150 ppm), we noticed a significant increase in curves that could not be confidently classified, but even in this case, our model identified 20 potential exocomet candidates. These promising results within Sector 1 motivate us to extend our analysis across all TESS sectors to detect and study comet-like activity in the extrasolar planetary systems.

astro-ph.EP↗

arXiv2022

Machine learning technique for morphological classification of galaxies from the SDSS. III. Image-based inference of detailed features

This paper follows series of our works on the applicability of various machine learning methods to the morphological galaxy classification (Vavilova et al., 2021, 2022). We exploited the sample of 315776 SDSS DR9 galaxies with absolute stellar magnitudes of -24m<Mr<-19.4m at 0.003<z<0.1 as a target data set for the CNN classifier based on the DenseNet-201. Because it is tightly overlapped with the Galaxy Zoo 2 (GZ2) sample, we use these annotated data as the training data set to classify galaxies into 34 detailed features. In the presence of a pronounced difference of visual parameters between galaxies from the GZ2 training data set and galaxies without known morphological parameters, we applied novel procedures, which allowed us for the first time to get rid of this difference for smaller and fainter SDSS galaxies. We describe in detail the adversarial validation technique as well as how we managed the optimal train-test split of galaxies from the training data set. We have also found optimal galaxy image transformations to increase the classifier generalization ability. It can be considered as another way to improve the human bias for those galaxy images that had a poor vote classification in the GZ project. Such an approach, likely auto-immunization, when the CNN classifier trained on very good images is able to retrain bad images from the same homogeneous sample, can be considered co-planar to other methods of combating the human bias. The accuracy of CNN classifier is in the range of 83.3-99.4 percent depending on 32 features. As a result, for the first time, we assigned the detailed morphological classification for more than 140K low-redshift galaxies, especially at the fainter end. We accentuate on the typical problem points of galaxy CNN image classification from the astronomical point of view. The catalogs will be available through the VizieR.

astro-ph.GA↗

arXiv2022

Machine learning technique for morphological classification of galaxies from SDSS. II. The image-based morphological catalogs of galaxies at 0.02<z<0.1

We applied the image-based approach with a convolutional neural network model to the sample of low-redshifts galaxies with $-24^{m}<M_{r}<-19.4^{m}$ from the SDSS DR9. We divided it into two subsamples, SDSS DR9 galaxy dataset and Galaxy Zoo 2 (GZ2) dataset, considering them as the inference and training datasets, respectively. As a result, we created the morphological catalog of 315782 galaxies at 0.02<z<0.1, where morphological five classes and 34 detailed features (bar, rings, number of spiral arms, mergers, etc.) were first defined for 216148 galaxies (inference dataset) by the image-based CNN classifier. For the rest of galaxies the initial morphological classification was re-assigned as in the GZ2 project. Our method shows the promising performance of morphological classification attaining more 93 % of accuracy for five classes morphology prediction except the cigar-shaped (75 %) and completely rounded (83 %) galaxies. Main results are presented in the catalog of 19468 completely rounded, 27321 rounded in-between, 3235 cigar-shaped, 4099 edge-on, 18615 spiral, and 72738 general low-redshift galaxies of the studied SDSS sample. As for the classification of galaxies by their detailed structural morphological features, our CNN model gives the accuracy in the range of 92-99 % depending on features, a number of galaxies with the given feature in the inference dataset, and the galaxy image quality. We demonstrate that implication of the CNN model with adversarial validation and adversarial image data augmentation improves classification of smaller and fainter SDSS galaxies with $m_{r}$ <17.7.

astro-ph.GA↗

arXiv2021

Machine learning technique for morphological classification of galaxies from the SDSS. I. Photometry-based approach

Methods. We used different galaxy classification techniques: human labeling, multi-photometry diagrams, Naive Bayes, Logistic Regression, Support Vector Machine, Random Forest, k-Nearest Neighbors, and k-fold validation. Results. We present results of a binary automated morphological classification of galaxies conducted by human labeling, multiphotometry, and supervised Machine Learning methods. We applied its to the sample of galaxies from the SDSS DR9 with 0.02 < z < 0.1 and 24m < Mr < 19.4m. To study the classifier, we used absolute magnitudes: Mu, Mg, Mr , Mi, Mz, Mu-Mr , Mg-Mi, Mu-Mg, Mr-Mz, and inverse concentration index to the center R50/R90. Using the Support vector machine classifier and the data on color indices, absolute magnitudes, inverse concentration index of galaxies with visual morphological types, we were able to classify 316 031 galaxies from the SDSS DR9 with unknown morphological types. Conclusions. The methods of Support Vector Machine and Random Forest with Scikit-learn machine learning in Python provide the highest accuracy for the binary galaxy morphological classification: 96.4% correctly classified (96.1% early E and 96.9% late L types) and 95.5% correctly classified (96.7% early E and 92.8% late L types), respectively. Applying the Support Vector Machine for the sample of 316 031 galaxies from the SDSS DR9 at z < 0.1, we found 141 211 E and 174 820 L types among them.

astro-ph.GA↗