Parametric Taylor series based latent dynamics identification neural networks

Xinlei Lin, Dunhui Xiao

incompletemedium confidence

Category: Not specified
Journal tier: Strong Field
Processed: Sep 28, 2025, 12:56 AM
arXiv Links: Abstract ↗PDF ↗

Audit review

The paper’s KNN–IDW scheme, online pipeline, and empirical claims (≤2% L2 errors; 94.7× and 176.5× speed-ups) are consistent and mostly correct, but key reproducibility details are missing and there is at least one typo in the Euclidean distance formula. The candidate solution gives a precise algorithmic specification and a valid exact-hit proof, but does not produce the required numerical tables/plots or measured speed-ups due to missing assets, so it is also incomplete.

Referee report (LaTeX)

\textbf{Recommendation:} minor revisions

\textbf{Journal Tier:} strong field

\textbf{Justification:}

A clear and lightweight latent-dynamics-based ROM with KNN–IDW interpolation is presented and demonstrated on two canonical problems, including multiscale grids. The empirical performance is strong and the method is well-motivated. Minor but important details are absent for full reproducibility, and there appear to be small typos (notably in the distance formula and timing narration). These can be straightforwardly fixed, after which the paper would be a solid contribution.