Paper revises latent force ODE filters for stability and calibration
Original title: Stable and Calibrated Latent Force ODE Filters
The joint posterior is a reweighted conventional posterior, and state prior variance controls how observations inform the force.
The paper studies calibration and stability of approximate Gaussian filters that jointly estimate system state and unobserved forces. It shows the joint posterior over state and force is a reweighted conventional latent-force posterior. Stability depends on the prior variance of the system state, which controls information flow from observations to the latent force. An improved plug-and-play algorithm is demonstrated on simulated experimental settings.
- The task is estimating unobserved forces in mechanistic systems.
- Inference remains a single forward-backward filtering pass.
- The revised algorithm is intended to be stable and calibrated.
- Demonstrations use simulated settings only.
From the original article
1 min read at arxiv.orgEstimating unobserved forces in mechanistic systems is a central task of scientific machine learning. Approximate Gaussian filtering methods that synthesize mechanistic and empirical information have shown remarkable capacity in jointly identifying latent processes together with the system state in a single forward-backward pass.
By Paul Fischer · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.