.png){kind=icon}
.png){kind=icon}
Return
Standard candle based distance estimation with learning algorithms
.png){kind=icon}
OA
AI
DOI：10.1016/j.ascom.2025.101044.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Measuring distances to celestial objects, such as stars and galaxies, is essential to characterizing their physical properties, formation, and evolution, and provides fundamental constraints on the expansion rate of the Universe. In this work, we present a methodological study comparing several machine learning and deep learning approaches for predicting astrophysical parameters - such as parallax, astrometry-based luminosity, and distance - using Cepheids and RR Lyrae samples from Gaia DR3 catalogues as input. In parallel, we introduce a framework to exploit the historical archive of photographic plates from INAF-OATo. In this context we extract a catalogue of objects from the plates, then cross-match the output sources with the Gaia dataset, with the goal of extending light curves that could serve as additional input for the models. Preliminary results identify the Gaussian Process regressor as the best-performing model among those tested, and the Multi-Layer Perceptron (MLP) as the most promising deep learning approach. We further study the propagation of uncertainties, enabling us to incorporate them both into the models and the predictions. For the plate analysis, we chose an image of the LMC field with 43 cepheids in common with the Gaia catalogue as a first case study to validate our methodology.
Keywords:
Gaia DR3
Standard candles
Machine learning
Image processing
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
A
IF:
1.8
Papers:
90
Citations:
0