Learning Energy-Based Models by Diffusion Recovery Likelihood
Ruiqi Gao UCLA、Yang Song Stanford University、Ben Poole Google Brain \ANDYing Nian Wu UCLA、Diederik P. Kingma Google Brain
Abstract
While energy-based models (EBMs) exhibit a number of desirable properties, training and sampling on high-dimensional datasets remains challenging. Inspired by recent progress on diffusion probabilistic models, we present a diffusion recovery likelihood method to tractably learn and sample from a sequence of EBMs trained on increasingly noisy versions of a dataset. Each EBM is trained with recovery likelihood, which maximizes the conditional probability of the data at a certain noise level given their noisy versions at a higher noise level. Optimizing recovery likelihood is more tractable than marginal likelihood, as sampling from the conditional distributions is much easier than sampling from the marginal distributions. After training, synthesized images can be generated by the sampling process that initializes from Gaussian white noise distribution and progressively samples the conditional distributions at decreasingly lower noise levels. Our method generates high fidelity samples on various image datasets. On unconditional CIFAR-10 our method achieves FID 9.58 and inception score 8.30, superior to the majority of GANs. Moreover, we demonstrate that unlike previous work on EBMs, o
中文速览
训练基于能量的生成模型（Energy-Based Model, EBM）时，最大似然估计需要从高维、多峰分布中做MCMC采样，极为困难，导致采样链不收敛、生成质量差。受扩散概率模型启发，本文提出"扩散恢复似然"（diffusion recovery likelihood）方法：先用扩散过程把数据逐步加噪构成一系列噪声版本，再针对每个噪声级别训练一个EBM，但训练目标换成"恢复似然"——即给定高噪声样本、预测低噪声样本的条件概率，这比直接优化边际似然容易得多，因为条件分布更集中、更容易采样。生成时从高斯白噪声出发，沿噪声级别从高到低依次做条件采样，最终得到清晰图像。实验显示，该方法在CIFAR-10无条件生成上达到FID 9.58、Inception Score 8.30，超越绝大多数GAN，同时长程MCMC链始终保持真实图像质量，使得精确估计归一化密度成为可能，解决了以往EBM长链采样退化的痼疾。
原文 arXiv:2012.08125;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2012.08125v2