The following is an explanation of some of the math in the paper "Variational auto-encoders with Student’s t-prior" by Najmeh Abiri1 and Mattias Ohlsson. In particular, I focused on their use of the reparametrization trick in order to train a VA with Student-t distribution instead of a Gaussian distribution.
URL/DOI link: VARIATIONAL AUTO-ENCODERS WITH STUDENT‘S T PRIOR EXCERPT
