Search

This work is part of the applications of neural networks in the study and modeling of the phenomena present in the solar photosphere. The proposed research is based on the network model generative adversaries making use of Pytorch's artificial intelligence modules. Wanted to train a neural network capable of generating groups of images of a high similarity with training images, These images correspond to physical magnitudes of the solar photosphere such as density, field magnetic, plasma speed, temperature, among others, obtained from the MURaM simulation code, although the neural network can be trained to generate images of any physical magnitude. work will focuses on magnetic field imaging in the solar photosphere. The results are presented training of the neural network, the comparison between the training images and the images generated, and the challenges to use these tools in the study of the solar photosphere are proposed, it is achieved demonstrate that GAN-type networks manage to partially recreate physical structures of the solar photosphere, with divergence values close to zero, which allows us to conclusively state that the structure created by the network is in technical terms, physically correct for the laws of magnetism, the following step in the investigation should focus on improving the network so that it can accurately recreate total structures of the solar photosphere that in turn allow evidence of a physical consistency with the MHD equations and have a higher resolution than the one generated in the current project. [ABSTRACT FROM AUTHOR]