E²-CRF accelerates frequency domain diffusion models by approximately 2.2 times via adaptive caching of stable high-frequency features triggered by residual error dynamics while preserving sample quality on five datasets.
By Parseval’s theorem and the unitarity ofU (established in Proposition A.10), we have Etotal =∥x∥ 2
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Accelerating Frequency Domain Diffusion Models with Error-Feedback Event-Driven Caching
E²-CRF accelerates frequency domain diffusion models by approximately 2.2 times via adaptive caching of stable high-frequency features triggered by residual error dynamics while preserving sample quality on five datasets.