Discrete Auto-regressive Variational Attention Models for Text Modeling

TL;DR

This paper introduces DAVAM, a novel discrete auto-regressive variational attention model for text modeling that effectively enriches the latent space and avoids posterior collapse, outperforming existing VAEs.

Contribution

The paper proposes a new auto-regressive variational attention mechanism with discrete latent space, addressing information underrepresentation and posterior collapse in VAEs.

Findings

DAVAM outperforms several VAE models on language modeling tasks.

The model effectively captures semantic dependencies in text.

It is mathematically proven to be free from posterior collapse.

Abstract

Variational autoencoders (VAEs) have been widely applied for text modeling. In practice, however, they are troubled by two challenges: information underrepresentation and posterior collapse. The former arises as only the last hidden state of LSTM encoder is transformed into the latent space, which is generally insufficient to summarize the data. The latter is a long-standing problem during the training of VAEs as the optimization is trapped to a disastrous local optimum. In this paper, we propose Discrete Auto-regressive Variational Attention Model (DAVAM) to address the challenges. Specifically, we introduce an auto-regressive variational attention approach to enrich the latent space by effectively capturing the semantic dependency from the input. We further design discrete latent space for the variational attention and mathematically show that our model is free from posterior collapse. Extensive experiments on language modeling tasks demonstrate the superiority of DAVAM against several VAE counterparts.