A new look on diffusion times for score-based generative models

Franzese, Giulio; Rossi, Simone; Yang, Lixuan; Finamore, Alessandro; Rossi, Dario; Filippone, Maurizio; Michiardi, Pietro
ICML 2022, 39th International Conference on Machine Learning, Continuous time methods for machine learning Workshop, 22-23 July 2022, Baltimore, USA

Score-based diffusion models are a class of generative models whose dynamics is described by stochastic differential equations that map noise into data. While recent works have started to lay down a theoretical foundation for these models, an analytical understanding of the role of the diffusion time T is still lacking. Current best practice advocates for a large T to ensure that the forward dynamics brings the diffusion sufficiently close to a known and simple noise distribution; however, a smaller value of T should be preferred for a better approximation of the score-matching objective and higher computational efficiency. Starting from a variational interpretation of diffusion models, in this work we quantify this trade-off, and suggest a new method to improve quality and efficiency of both training and sampling, by adopting smaller diffusion times. Indeed, we show how an auxiliary model can be used to bridge the gap between the ideal and the simulated forward dynamics, followed by a standard reverse diffusion process. Empirical results support our analysis; for image data, our method is competitive w.r.t. the state-of-the-art, according to standard sample quality metrics and log-likelihood.

 

HAL
Type:
Conférence
City:
Baltimore
Date:
2022-07-22
Department:
Data Science
Eurecom Ref:
6934
Copyright:
© 2022 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

PERMALINK : https://www.eurecom.fr/publication/6934