Spectral and Projection Methods
A class of posterior sampling methods that reduce the difficulty of conditioning diffusion models through low-rank approximation or projection operations. They may sacrifice some accuracy in exchange for computational efficiency.
Timeline (last 90 days)
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.
Spectral and projection methods reduce the difficulty of conditioning through low-rank approximation or projection operations, assuming the posterior can be captured by low-dimensional structure, possibly at the cost of some accuracy.
All Facts (2)
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.
50%UnverifiedSpectral and projection methods reduce the difficulty of conditioning through low-rank approximation or projection operations, assuming the posterior can be captured by low-dimensional structure, possibly at the cost of some accuracy.
50%