Estimating the Optimal Covariance with Imperfect Mean in Diffusion Probabilistic Models

06/15/2022
by   Fan Bao, et al.
0

Diffusion probabilistic models (DPMs) are a class of powerful deep generative models (DGMs). Despite their success, the iterative generation process over the full timesteps is much less efficient than other DGMs such as GANs. Thus, the generation performance on a subset of timesteps is crucial, which is greatly influenced by the covariance design in DPMs. In this work, we consider diagonal and full covariances to improve the expressive power of DPMs. We derive the optimal result for such covariances, and then correct it when the mean of DPMs is imperfect. Both the optimal and the corrected ones can be decomposed into terms of conditional expectations over functions of noise. Building upon it, we propose to estimate the optimal covariance and its correction given imperfect mean by learning these conditional expectations. Our method can be applied to DPMs with both discrete and continuous timesteps. We consider the diagonal covariance in our implementation for computational efficiency. For an efficient practical implementation, we adopt a parameter sharing scheme and a two-stage training process. Empirically, our method outperforms a wide variety of covariance design on likelihood results, and improves the sample quality especially on a small number of timesteps.

READ FULL TEXT

page 26

page 27

page 28

page 29

page 30

research
05/04/2023

LayoutDM: Transformer-based Diffusion Model for Layout Generation

Automatic layout generation that can synthesize high-quality layouts is ...
research
05/30/2022

Few-Shot Diffusion Models

Denoising diffusion probabilistic models (DDPM) are powerful hierarchica...
research
03/20/2023

Practical Realization of Bessel's Correction for a Bias-Free Estimation of the Auto-Covariance and the Cross-Covariance Functions

To derive the auto-covariance function from a sampled and time-limited s...
research
01/17/2022

Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models

Diffusion probabilistic models (DPMs) represent a class of powerful gene...
research
05/26/2023

High-Fidelity Image Compression with Score-based Generative Models

Despite the tremendous success of diffusion generative models in text-to...
research
09/15/2023

Denoising Diffusion Probabilistic Models for Hardware-Impaired Communications

Generative AI has received significant attention among a spectrum of div...

Please sign up or login with your details

Forgot password? Click here to reset