Back to search
2505.00846

ON THE EMERGENCE OF NUMERICAL INSTABILITIES IN NEXT GENERATION RESERVOIR COMPUTING

Edmilson Roque dos Santos, Erik Bollt

wronghigh confidence
Category
math.DS
Journal tier
Specialist/Solid
Processed
Sep 28, 2025, 12:56 AM

Audit review

The paper states as a formal Result that, for small time lag τ and step size h, the NGRC feature matrix Ψ is rank deficient for any delay dimension k>1, and its proof asserts linear dependence of certain columns; however, the provided argument actually shows only that two columns become O(h)-close via a mean-value/Euler step and then appeals to continuity of singular values to claim small σ_min, not σ_min=0. In short, the paper’s proof demonstrates ill-conditioning (near-dependence), not exact rank deficiency for any fixed h>0, while the model’s solution correctly quantifies σ_min(Ψ) ≤ C·τh and explains that exact dependence requires extra structure on F and is not generic. See Section 5.2 and the displayed equations and column constructions in the paper’s proof for details .

Referee report (LaTeX)

\textbf{Recommendation:} major revisions

\textbf{Journal Tier:} specialist/solid

\textbf{Justification:}

The paper provides practically important numerical analysis of NGRC conditioning and ties it convincingly to hyperparameter choices. However, its formal claim of rank deficiency for small τ and h overreaches the proof provided, which establishes only near-dependence/ill-conditioning. Clarifying this to a quantitative ill-conditioning result and specifying conditions for exact rank loss are necessary to ensure correctness. With these revisions, the paper would constitute a solid contribution for specialists.