Diffusion Priors In Variational Autoencoders

Abstract

Among likelihood-based approaches for deep generative modelling, variational autoencoders (VAEs) offer scalable amortized posterior inference and fast sampling. However, VAEs are also more and more outperformed by competing models such as normalizing flows (NFs), deep-energy models, or the new denoising diffusion probabilistic models (DDPMs). In this preliminary work, we improve VAEs by demonstrating how DDPMs can be used for modelling the prior distribution of the latent variables. The diffusion prior model improves upon Gaussian priors of classical VAEs and is competitive with NF-based priors. Finally, we hypothesize that hierarchical VAEs could similarly benefit from the enhanced capacity of diffusion priors.

References in corpus (12)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Denoising Diffusion Probabilistic Models
• Diffusion Models Beat GANs on Image Synthesis
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Variational Inference with Normalizing Flows
• Improved Denoising Diffusion Probabilistic Models
• NVAE: A Deep Hierarchical Variational Autoencoder
• Variational Lossy Autoencoder
• Video Compression With Rate-Distortion Autoencoders
• Improving Variational Inference with Inverse Autoregressive Flow
• Isolating Sources of Disentanglement in Variational Autoencoders
• Score-based Generative Modeling in Latent Space

Cited by in corpus (2)

• Deep Generative Models for Detector Signature Simulation: A Taxonomic Review
• Score-based Generative Modeling in Latent Space