Tag
The paper introduces a divergence-kernel method for jointly training neural SDE drift and diffusion fields with losses that depend directly on the terminal density, achieving fewer off-mode samples than score-based generators on Gaussian-mixture benchmarks with time-independent fields.