Skip to content

Latest commit

 

History

History
21 lines (17 loc) · 800 Bytes

README.md

File metadata and controls

21 lines (17 loc) · 800 Bytes
Raw

Nested Dropout

Implementation of nested dropout (https://arxiv.org/abs/1402.0915), but modernized. Uses Gumbel-Bernoulli distribution for the latent vector.

Generates an ordered code for the latent space of the autoencoder, this has many uses e.g. adaptive compression.

img 0 img 1

Requirements

  • tensorflow
  • tensorflow_probability
  • numpy
  • matplotlib

Usage

First pretrain a model with the pretrain_model.py script (should only take a few minutes). Then feed this pretrain model to the train_model.py script (this will take awhile).

TODO

  • Add KL term so it's similar to a VAE
  • code clean up