Bayesian Neural Network for inferring morphological parameters of Low Surface Brightness Galaxy

In this thesis, we aim to filter out biases of Bayesian Neural Network trained on synthetically generated images of Low Surface Brightness Galaxies.

Low Surface Brightness Galaxies (LSBGs) are galaxies with central brightness at least a magnitude fainter than that of the ambient sky. They are of interest to reseachers as their number are expected to dominate the galaxy population. Unfortunately, because of their low signal-to-noise nature, understanding the properties or obtaining the images of LSBGs are no easy feats. Bayesian Neural Network can be trained on synthetically generated galaxy images in order to achieve such task. However, the posterior distribution outputted by the Bayesian Neural Network then would be biased by the synthetic distribution. In this work, we incorporate a hierarchical inference step in order to filter out such biases. Furthermore, we will explore other appraches in inferring morphological parameters of LSBGs. Recording: https://drexel.zoom.us/rec/share/a1eb3YlON8e1o_dCK23cQvfe_Km1vG6xECF7TL58R6CUI4tY3lbXxu7NORiQnk4g.BhPdvJ8ja7ju_JDM Passcode: m#M.u2y1

Team Members

Advisors