Convergence analysis and application for high-order neural networks based on gradient descent learning algorithm via smooth regularization
Khidir Shaib Mohamed, Alawia Adam, Yousif Shoaib Mohammed, Yan Xiong · 2025
Read Convergence analysis and application for high-order neural networks based on gradient descent learning algorithm via smooth regularization by Khidir Shaib Mohamed, Alawia Adam, Yousif Shoaib Mohammed, Yan Xiong (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/pmc-12770431.txt.
Description
Some neural networks multiply their inputs together, gaining speed and a feel for tangled patterns, but training them smoothly is tricky. Khidir Shaib Mohamed, Alawia Adam, Yousif Shoaib Mohammed, and Yan Xiong propose a training method built on gradient descent with a penalty that prunes weak links, in this 2025 paper. The raw penalty misbehaves at zero, causing wobbles and hard math, so they smooth it first.
The smoothed method trains the high-order pi-sigma network in batches. On a classic parity puzzle and a nonlinear signal task, it generalized better and pruned more cleanly than four rival penalty methods. The authors prove the training settles down in both strong and weak senses, and that its error falls steadily. The result is a steadier recipe for a fast but finicky network family.
Catalog record
| CREATOR | Khidir Shaib Mohamed, Alawia Adam, Yousif Shoaib Mohammed, Yan Xiong |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |