Divergence-Kernel method for scores of random systems

TL;DR

This paper introduces a divergence-kernel formula for scores of random dynamical systems, applicable to continuous-time SDEs with multiplicative noise, and demonstrates a Monte-Carlo algorithm on high-dimensional systems.

Contribution

It presents a novel divergence-kernel formula for scores in random systems, including special cases and a practical Monte-Carlo algorithm for high-dimensional systems.

Findings

Derived divergence-kernel formula applicable to SDEs with multiplicative noise

Developed a pathwise Monte-Carlo algorithm for score estimation

Demonstrated effectiveness on 40-dimensional Lorenz 96 system

Abstract

We derive the divergence-kernel formula for the scores of random dynamical systems, then formally pass to the continuous-time limit of SDEs. Our formula works for multiplicative noise systems over any period of time; it does not require hyperbolicity. We also consider several special cases: (1) for additive noise, we give a pure kernel formula; (2) for short-time, we give a pure divergence formula; (3) we give a formula which does not involve scores of the initial distribution. Based on the new formula, we derive a pathwise Monte-Carlo algorithm for scores, and demonstrate it on the 40-dimensional Lorenz 96 system with multiplicative noise.