Augmented Normalizing Flows: Bridging the Gap Between Generative Flows and Latent Variable Models

TL;DR

This paper introduces augmented normalizing flows that enhance expressivity in generative modeling by bridging flows and latent variable models, achieving state-of-the-art results efficiently.

Contribution

It proposes a new family of augmented flows that approximate Hamiltonian ODEs, improving expressivity without high computational costs.

Findings

Achieves state-of-the-art performance on standard benchmarks

Proves the flow can approximate Hamiltonian ODEs as universal transport maps

Enhances expressivity of generative flows efficiently

Abstract

In this work, we propose a new family of generative flows on an augmented data space, with an aim to improve expressivity without drastically increasing the computational cost of sampling and evaluation of a lower bound on the likelihood. Theoretically, we prove the proposed flow can approximate a Hamiltonian ODE as a universal transport map. Empirically, we demonstrate state-of-the-art performance on standard benchmarks of flow-based generative modeling.