mu-Forcing: Training Variational Recurrent Autoencoders for Text Generation

Introduction

In this work, we explore the reason why training Variational Recurrent Autoencoders (VRAE) for text generation suffers from serious uninformative latent variables problem and propose an effective regularizer based approach to address it.

Code Link

code

Paper

µ-Forcing: Training Variational Recurrent Autoencoders for Text Generation, TALLIP 2019 Paper