Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design
Jonathan Ho Xi Chen Aravind Srinivas Yan Duan Pieter Abbeel
Abstract
Flow-based generative models are powerful exact likelihood models with efficient sampling and inference. Despite their computational efficiency, flow-based models generally have much worse density modeling performance compared to state-of-the-art autoregressive models. In this paper, we investigate and improve upon three limiting design choices employed by flow-based models in prior work: the use of uniform noise for dequantization, the use of inexpressive affine flows, and the use of purely convolutional conditioning networks in coupling layers. Based on our findings, we propose Flow++, a new flow-based model that is now the state-of-the-art non-autoregressive model for unconditional density estimation on standard image benchmarks. Our work has begun to close the significant performance gap that has so far existed between autoregressive models and flow-based models. Our implementation is available at: https://github.com/aravindsrinivas/flowpp.
中文速览
流模型（flow-based model）在生成图像时推理和采样都很快，却一直在密度估计精度上远输给自回归模型。研究者找到了三个根本性短板：用均匀噪声做离散数据的连续化处理（dequantization）会迫使模型不自然地铺平密度、仿射耦合层表达能力太弱、耦合层里的条件网络只靠卷积而缺乏全局感知能力。针对这三点，他们提出 Flow++：用基于变分推断的流式去量化代替均匀噪声，用逻辑斯谛混合分布的 CDF 构造更灵活的耦合变换，并在条件网络中引入多头自注意力机制。实验表明，Flow++ 在 CIFAR-10 和 ImageNet 的无条件密度估计上刷新了非自回归模型的最优成绩，显著缩小了与自回归模型之间长期存在的性能差距，为兼具高质量密度建模和快速采样的生成模型提供了新的可行路径。
原文 arXiv:1902.00275;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1902.00275v2