Towards Neural Kolmogorov Equations: Parallelizable SDE Learning with Neural PDEs

Arthur Bizzi, Olga Fink

Keywords: Neural SDEs, Fokker-Planck, Kolmogorov, Gaussian Mixtures

Abstract: The presence of stochasticity makes learning differential equations from data substantially harder, requiring Neural SDEs to be trained with costly procedures involving repeated sequential integration. We introduce Neural Kolmogorov Equations, a parallelizable framework for learning continuous stochastic processes, based on the deterministic framework of the Forward-Kolmogorov Equation.

Serve As Reviewer: ~Arthur_Bizzi1

Submission Number: 41