Skip to content

Fisher Information and Dynamical Sampling I

Home / Papers / Fisher Information and Dynamical Sampling I

Fisher Information and Dynamical Sampling I

Authors: Mattia Carrino, Stefan Hohenegger Date: 2026-04-27 Paper ID: openalex:2604.24499

Summary

This paper investigates the information-theoretic limitations of reconstructing dynamical system trajectories from finite time-series measurements. The authors derive the asymptotic bias of Fisher information for these reconstructions, establishing a quantitative link between measurement size and trajectory estimation accuracy. Furthermore, they demonstrate that clustering degrees of freedom effectively mitigates this bias, offering a principled approach to evaluating information loss in reduced-order models of complex dynamical systems.

Key Contributions

  • Derives the large-sample bias of Fisher information for reconstructing dynamical system trajectories from discrete measurements.
  • Demonstrates that clustering degrees of freedom systematically reduces reconstruction bias and improves estimation accuracy.
  • Provides a quantitative framework for assessing information loss in dimensionality-reduced dynamical models.

Open Questions & Future Work

Archivist Review

The paper provides a rigorous information-theoretic analysis of dynamical system reconstruction bias. I approved the open question regarding the optimal trade-off between clustering-induced bias reduction and information loss, as this is a fundamental bottleneck in high-dimensional time-series modeling. Other candidates were rejected as they were either too specific to the paper’s analytical derivation or lacked the generality to serve as vault-level concepts.

Approved Open Questions

  • Optimal Clustering for Dynamical Sampling: Understanding this trade-off is critical for applying information-theoretic analysis to real-world dynamical systems where data is often limited, noisy, and high-dimensional.

Rejected Candidates

  • [concept] Fisher Information Bias for Dynamical Sampling (fisher-information-bias-estimation) - not_reusable: This is a theoretical analysis of a specific estimator rather than a general methodological concept suitable for the vault.
  • [concept] Information Loss Framework for Dimensionality Reduction (information-loss-quantification-framework) - not_novel: This describes the application of Fisher information in a specific context; it is not a distinct architectural or algorithmic framework that would recur in future papers.

Metadata & Links

url
https://arxiv.org/abs/2604.24499
paper_id
2604.24499
paper_source
openalex
domain
time-series
tags
architectures
datasets
concept_slugs
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-04-30T06:04:08Z