RRT-VAE

Pytorch implementation for the paper: Learning VAE-LDA Models with Rounded Reparameterization Trick

To run the code

python RRT.py

Tunable parameters:

-t  # default=50 #number of topics
-e  # default=150 # number of training epochs
-b  # default=100 # training batch size
-l  # default=0.001 # learning rate
-a  # default=1.0 # Dirichlet prior parameter
-d  #default=1e10 # Delta parameter for RRT
-lam # default=0.01 # Lambda parameter for RRT