Adversarially Regularized Autoencoders

TL;DR

This paper introduces an adversarially regularized autoencoder (ARAE) framework for modeling discrete data like text sequences, enabling natural language generation, style transfer, and controllable representation learning.

Contribution

It extends Wasserstein autoencoders to discrete structures and demonstrates effective text generation and style transfer with improved evaluation metrics.

Findings

Effective generation of natural textual outputs

Successful unaligned textual style transfer

Enhanced latent space manipulation capabilities

Abstract

Deep latent variable models, trained using variational autoencoders or generative adversarial networks, are now a key technique for representation learning of continuous structures. However, applying similar methods to discrete structures, such as text sequences or discretized images, has proven to be more challenging. In this work, we propose a flexible method for training deep latent variable models of discrete structures. Our approach is based on the recently-proposed Wasserstein autoencoder (WAE) which formalizes the adversarial autoencoder (AAE) as an optimal transport problem. We first extend this framework to model discrete sequences, and then further explore different learned priors targeting a controllable representation. This adversarially regularized autoencoder (ARAE) allows us to generate natural textual outputs as well as perform manipulations in the latent space to induce change in the output space. Finally we show that the latent representation can be trained to perform unaligned textual style transfer, giving improvements both in automatic/human evaluation compared to existing methods.