Coarse-grained sensitivity for multiscale data assimilation

TL;DR

This paper introduces a novel approach using the effective average action and its gradient to improve multiscale data assimilation, enabling better handling of slow and fast degrees of freedom.

Contribution

It presents a new method for numerically evaluating the gradient of the effective average action to solve multiscale data assimilation problems.

Findings

Effective average action aids in multiscale data assimilation.

Numerical procedure for gradient evaluation is demonstrated.

Improved solution of variational problems for slow variables.

Abstract

We show that the effective average action and its gradient are useful for solving multiscale data assimilation problems. We also present a procedure for numerically evaluating the gradient of the effective average action, and demonstrate that the variational problem for slow degrees of freedom can be solved properly using the "effective gradient."