Discrete Autoencoders for Sequence Models
Abstract
The contributions of this paper are:
* A discretization technique that works well without any extra losses or parameters to tune.
* A way to measure performance of autoencoders for sequence models (and baselines).
* An improved way to sample from sequence models (trained with an autoencoder part).
* A discretization technique that works well without any extra losses or parameters to tune.
* A way to measure performance of autoencoders for sequence models (and baselines).
* An improved way to sample from sequence models (trained with an autoencoder part).
Research Areas
