Attractor Geometry Limits System Discovery in Chaotic Dynamics.

Matteo Gallo, Fabio Anselmi, Paolo Lazzari· July 22, 2026 View original

Summary

This research shows that the "attractor geometry" of a system, measured by the smallest eigenvalue of its invariant-measure moment matrix, sets a fundamental limit on how well governing equations can be discovered from data. It demonstrates that this metric predicts the identifiability ceiling for both sparse regression and evolutionary symbolic regression algorithms.

New research explores the fundamental limits of symbolically discovering governing equations from data, beyond just algorithm design or data volume. The study introduces the concept of "attractor geometry," specifically the smallest eigenvalue of the invariant-measure moment matrix, as a key factor. This metric, derived from the Birkhoff ergodic theorem, can be calculated from a short reference trajectory and quantifies how thoroughly an attractor covers function space. The findings indicate that where this metric approaches zero, recovering equations becomes impossible for any algorithm, while higher values correlate with improved algorithm performance. Interestingly, while chaos generally enhances this metric by spreading the attractor, it also amplifies noise, which can affect different discovery methods in opposing ways. The research also proposes a new structural metric, Soft F1, to better assess performance differences.

Why it matters

Professionals developing or applying AI for system identification and modeling need to understand these inherent limitations to set realistic expectations and choose appropriate methodologies. This work provides a theoretical framework to predict when system discovery will be feasible and how different algorithms might perform.

How to implement this in your domain

  1. 1Evaluate the $\lambda_{\min}(M)$ metric for your system's data before attempting symbolic regression to assess identifiability.
  2. 2Consider the impact of system chaos and noise amplification when selecting between sparse regression (SINDy) and evolutionary symbolic regression (PySR).
  3. 3Integrate the Soft F1 metric into your evaluation pipeline for a more nuanced assessment of model structural accuracy.
  4. 4Prioritize data collection strategies that maximize attractor coverage in function space to improve system discovery potential.

Who benefits

Scientific ResearchEngineeringAerospaceClimate ModelingFinance

Key takeaways

  • The geometry of a system's attractor fundamentally limits the identifiability of its governing equations.
  • A new metric, $\lambda_{\min}(M)$, can predict the ceiling for system discovery algorithms.
  • Chaos can improve identifiability by spreading the attractor but also amplifies noise, affecting algorithms differently.
  • A new Soft F1 metric offers a more precise way to evaluate structural performance in symbolic regression.

Original post by Matteo Gallo, Fabio Anselmi, Paolo Lazzari

"arXiv:2607.18490v1 Announce Type: new Abstract: Symbolic discovery of governing equations from data is limited not only by algorithm design and data volume, but by the geometry of the attractor: what the long-run dynamics allow to be recovered. Using a within-system design on Lor…"

View on X

Originally posted by Matteo Gallo, Fabio Anselmi, Paolo Lazzari on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI Engineering & DevToolsAI Research

New Tool Generates Contamination-Resistant, Labeled Code Datasets for LLMs

Spaghetti Architect is a new open-source tool that generates controlled, multi-language code datasets, addressing issues of contamination and lack of semantic control in existing code corpora. It creates correct-by-construction programs with adjustable "messiness" and difficulty labels, making it ideal for training and evaluating code-generating LLMs.

Yuxiang JiJul 22, 2026
AI ResearchAI Engineering & DevTools

New Method Safely Gates Hazardous LLM Knowledge Without Deletion

Researchers introduce Token Inoculation, a method that allows large language models to retain sensitive "dual-use" knowledge while selectively refusing hazardous queries. This approach uses a special token to condition the model's behavior, improving safety without sacrificing benign domain performance.

Seunghyun Lee, Dongyoon Han, Sangdoo YunJul 22, 2026
AI ResearchAI Engineering & DevTools

GNNAS-TSP Selects Optimal Algorithms for Traveling Salesman Problem

Researchers introduce GNNAS-TSP, a Graph Neural Network (GNN)-based framework for automated algorithm selection (AS) for the Traveling Salesman Problem (TSP). GNNAS-TSP learns TSP instance representations directly from raw graph data, avoiding manual feature engineering, and formulates AS as a joint cost-prediction and ranking task to select the best solver from a portfolio under fixed computational budgets.

Zhaoxuan Li, Jiale Yang, Yifei Lu, Mustafa MisirJul 22, 2026