Una red neuronal para la búsqueda de exoplanetas utilizando el método de velocidad radial

NIETO, L. A.; DÍAZ, R. F.; SEGURA, E. C.

Congreso; 62a Reunión Anual Asociación Argentina de Astronomía; 2020

The search for exoplanets is a field that strongly requires the observation and analysis of a large amount of data. Improvements in techniques and instruments made it possible for planets with increasingly smaller masses to be discovered using radial velocity and transit methods; but it was only recently that the planets with the smallest mass and radius of the planet population began to be probed in detail. This was largely due to the launch of the Kepler mission, and the improvement in the methods used to extract information from radial velocity data. At present, missions such as GAIA and TESS provide an increasingly large amount of information, and that is why the community is looking towards data science and the different artificial intelligence methodologies, as an important support in the face of this avalanche of data. Already in recent years, studies have begun to appear in the field of extrasolar planets that make use of these techniques. The objective of this work is to explore the scope of neural networks in signal analysis and look for mechanisms to complement or replace current methods. For this, synthetic data were generated that seek to imitate radial velocity measurements of solar-type stars, and a neural network was designed and trained to classify them. The results were compared with the most common method to detect planets in time series, and it was shown that the neural network achieves a decrease of 28 % in false positives with an improvement of five orders of magnitude in the execution time.