ConceptSDE匹配
SDE Matching
A class of methods for training neural SDEs by matching the statistical properties of real data trajectories to fit the drift and diffusion parameters of an SDE, without requiring stepwise numerical integration. The approach shares conceptual connections with denoising score matching in diffusion models.
Timeline (last 90 days)
Oct 5
SDECast extends the SDE Matching method to enable direct learning of stochastic dynamics in physical space.
Unverified50%