MaCow: Masked Convolutional Generative Flow

TL;DR

MaCow introduces a masked convolutional flow architecture that enhances density estimation performance and training efficiency, narrowing the gap with autoregressive models in image generation tasks.

Contribution

The paper proposes MaCow, a novel masked convolutional flow model that improves density estimation and training stability over existing flow-based models.

Findings

MaCow outperforms Glow on standard image benchmarks.

MaCow achieves faster and more stable training.

MaCow narrows the performance gap with autoregressive models.

Abstract

Flow-based generative models, conceptually attractive due to tractability of both the exact log-likelihood computation and latent-variable inference, and efficiency of both training and sampling, has led to a number of impressive empirical successes and spawned many advanced variants and theoretical investigations. Despite their computational efficiency, the density estimation performance of flow-based generative models significantly falls behind those of state-of-the-art autoregressive models. In this work, we introduce masked convolutional generative flow (MaCow), a simple yet effective architecture of generative flow using masked convolution. By restricting the local connectivity in a small kernel, MaCow enjoys the properties of fast and stable training, and efficient sampling, while achieving significant improvements over Glow for density estimation on standard image benchmarks, considerably narrowing the gap to autoregressive models.