DynaTab: Dynamic Feature Ordering as Neural Rewiring for High-Dimensional Tabular Data
Al Zadid Sultan Bin Habib, Gianfranco Doretto, Donald A. Adjeroh · 2026
Read DynaTab: Dynamic Feature Ordering as Neural Rewiring for High-Dimensional Tabular Data by Al Zadid Sultan Bin Habib, Gianfranco Doretto, Donald A. Adjeroh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-03430.txt.
Description
Al Zadid Sultan Bin Habib, Gianfranco Doretto, and Donald A. Adjeroh present DynaTab, a deep architecture for high-dimensional tabular data. It learns the best feature order for a dataset. Tables have no natural column order. That blocks sequence-sensitive deep models, which expect a meaningful ordering.
Inspired by neural rewiring, DynaTab reorders features dynamically. Learned position embeddings, importance gating, and masked attention process the result, fitting any sequence backbone. A lightweight criterion predicts when reordering will help, based on dataset complexity. Against dozens of baselines on many real datasets, gains were significant. They were largest in high dimensions, supporting ordering as a new design lever.
Catalog record
| CREATOR | Al Zadid Sultan Bin Habib, Gianfranco Doretto, Donald A. Adjeroh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 149 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.03430