Autoregressive Quantile Networks for Generative Modeling

TL;DR

This paper presents AIQN, a novel autoregressive implicit quantile network approach for generative modeling that improves perceptual quality and evaluation metrics without sacrificing diversity, applicable to models like PixelCNN.

Contribution

The paper introduces AIQN, a new quantile regression-based method for generative modeling, extending PixelCNN and demonstrating superior perceptual quality and stability.

Findings

AIQN improves perceptual quality and evaluation metrics.

AIQN maintains high sample diversity.

AIQN is stable and versatile across architectures.

Abstract

We introduce autoregressive implicit quantile networks (AIQN), a fundamentally different approach to generative modeling than those commonly used, that implicitly captures the distribution using quantile regression. AIQN is able to achieve superior perceptual quality and improvements in evaluation metrics, without incurring a loss of sample diversity. The method can be applied to many existing models and architectures. In this work we extend the PixelCNN model with AIQN and demonstrate results on CIFAR-10 and ImageNet using Inception score, FID, non-cherry-picked samples, and inpainting results. We consistently observe that AIQN yields a highly stable algorithm that improves perceptual quality while maintaining a highly diverse distribution.