Paperlane

no. 6 · Monday 2 November 2026

Generative SDEs with Trainable Diffusion and Density-Based Losses

Angxiu Ni

math.DS · via cs.LG · 17 pp

Abstract

We develop a divergence-kernel method for jointly optimizing neural SDE drift and diffusion fields and the initial Gaussian distribution, using losses that depend directly on the terminal density. In two- and eight-dimensional Gaussian-mixture experiments, we learn time-independent fields and obtain far fewer off-mode samples than score-based generators with similar parameter counts. The drift attracts samples toward the modes, while the diffusion helps disperse excess probability between them.

comments
17 pages, 5 figures
fields
Dynamical Systems
pdf ↗ page arXiv:2610.41243 ↗