Using information-theoretic identities, the authors prove non-asymptotic KL bounds for discrete-time diffusion sampling and show that matching higher moments accelerates convergence.
Mutual information and minimum mean-square error in Gaussian channels,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
stat.ML 1years
2025 1verdicts
ACCEPT 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Information-Theoretic Proofs for Diffusion Sampling
Using information-theoretic identities, the authors prove non-asymptotic KL bounds for discrete-time diffusion sampling and show that matching higher moments accelerates convergence.