Variable Length Embeddings

Johnathan Chiu, Andi Gu, Matt Zhou

In this work, we introduce a novel deep learning architecture, Variable Length Embeddings (VLEs), an autoregressive model that can produce a latent representation composed of an ar…