Abstract
Magnetic fields are believed to play a crucial role in stellar evolution. To better understand this evolution, it is essential to measure the magnetic fields on the stellar surface. These measurements can be achieved through spectropolarimetric observations, using the polarized radiative transfer equation. Magnetic field properties are inferred by adjusting the Stokes profiles. In this study, we propose a deep learning approach using a feed-forward neural network to estimate the Stokes profiles based on eight input parameters that describe the magnetic field configuration. To achieve this, we conducted scaling experiments on the data, explored different configurations of the FNN architecture, and compared two approaches. A model capable of accurately estimating the Stokes profiles I, Q and V was obtained. However, we encountered difficulties in estimating Stokes profiles Q and U when they have low amplitudes.
