ConceptScore Matching
Score Matching
A statistical method for learning generative models by estimating the gradient field of the log-density of the data distribution, known as the score. It is one of the mathematical foundations of diffusion models.
Timeline (last 90 days)
Oct 16
StochLOB proposes a training-only score matching framework that changes the corruption law of the forward diffusion, while the predictor at deployment time remains deterministic.
Unverified50%