Learning Energy-Based Models by Diffusion Recovery Likelihood
This work addresses the challenge of training and sampling high-dimensional EBMs, which is a problem for researchers and practitioners working with generative models, offering a more stable and tractable approach.
This paper introduces a diffusion recovery likelihood method to train and sample energy-based models (EBMs) on high-dimensional datasets by learning a sequence of EBMs on increasingly noisy data. The method achieves high-fidelity image generation, with an FID of 9.58 and an Inception Score of 8.30 on unconditional CIFAR-10, outperforming most GANs.
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, our long-run MCMC samples from the conditional distributions do not diverge and still represent realistic images, allowing us to accurately estimate the normalized density of data even for high-dimensional datasets. Our implementation is available at https://github.com/ruiqigao/recovery_likelihood.