Conceptamortized conditional training
Amortized Conditional Training
Offline training of a generative model conditioned on observations, moving most of the computational cost to the training stage. Sampling at online inference is very fast, making it suitable for repeated inference scenarios, though it depends on the training distribution.
Timeline (last 90 days)
Nov 3
The survey categorizes the main methods of diffusion model posterior sampling into four classes: likelihood guidance, spectral and projection methods, sequential Monte Carlo correction, and amortized conditional training.
Unverified50%