Enhancing Neural Function Approximation: The XNet Outperforming KAN
Xin Li, Xiaotao Zheng, Zhihong Xia · 2025
Read Enhancing Neural Function Approximation: The XNet Outperforming KAN by Xin Li, Xiaotao Zheng, Zhihong Xia (2025) 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-2501-18959.txt.
Description
Xin Li, Xiaotao Zheng, and Zhihong Xia introduce a one-layer network design that fits complex functions with high accuracy, in a paper from 2025. Standard networks gain accuracy by adding depth, and a recent rival design uses spline curves on each link. Both routes can be slow or imprecise for hard functions.
Their design uses activation curves drawn from a classic integral formula. Theory shows these curves can reach any chosen order of accuracy as they refine. In tests on function fitting, equation solving, and game learning, error fell by huge factors versus rivals and training ran much faster. The authors pitch it for science computing and AI alike.
Catalog record
| CREATOR | Xin Li, Xiaotao Zheng, Zhihong Xia |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-18959