On the Importance of the Kullback-Leibler Divergence Term in Variational Autoencoders for Text Generation

TL;DR

This paper investigates the role of the KL divergence term in Variational Autoencoders for text generation, demonstrating how explicit constraints can prevent posterior collapse and influence the information encoding and generative performance.

Contribution

It introduces an explicit constraint on the KL divergence in VAEs, providing insights into its importance for avoiding posterior collapse and balancing information encoding with generative capacity.

Findings

Explicit KL constraint prevents posterior collapse.

Trade-off between information encoding and generation quality.

Enhanced understanding of the KL term's role in VAE training.

Abstract

Variational Autoencoders (VAEs) are known to suffer from learning uninformative latent representation of the input due to issues such as approximated posterior collapse, or entanglement of the latent space. We impose an explicit constraint on the Kullback-Leibler (KL) divergence term inside the VAE objective function. While the explicit constraint naturally avoids posterior collapse, we use it to further understand the significance of the KL term in controlling the information transmitted through the VAE channel. Within this framework, we explore different properties of the estimated posterior distribution, and highlight the trade-off between the amount of information encoded in a latent code during training, and the generative capacity of the model.