Diffusion models can be trained up to roughly 4x faster in the paper's experiments by reducing trajectory miscibility via KNN noise selection or image scaling, though the mechanism is not fully isolated.
Conditional wasser- stein distances with applications in bayesian ot flow matching, 2024
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
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Improved Immiscible Diffusion: Accelerate Diffusion Training by Reducing Its Miscibility
Diffusion models can be trained up to roughly 4x faster in the paper's experiments by reducing trajectory miscibility via KNN noise selection or image scaling, though the mechanism is not fully isolated.