/VAE

VAE in the continuous and discrete versions

Primary LanguagePython

VAE implemented by Jiangchao Yao

VAE in the continuous and discrete versions (including the corresponding two kinds of importance-weighted measures respectively in the perspective of log-likelihood and KL-divergence.).


Reference paper:

[1] "Autoencoding variational bayes"

[2] "Gumbel-softmax for categorical reparameterization" or "The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables"