Kolmogorov GAM Networks are all you need!
Sarah Polson, Vadim Sokolov · 2025
Read Kolmogorov GAM Networks are all you need! by Sarah Polson, Vadim Sokolov (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-2501-00704.txt.
Description
Sarah Polson and Vadim Sokolov present a lean network design drawn from a classic theorem in a 2025 paper. Kolmogorov's superposition theorem says any multi-input function splits into sums of single-input parts. Their network turns that split into a trainable architecture.
The model pairs a fixed embedding, shared across tasks, with an additive layer. That layer follows the generalized additive model style that statisticians trust for clarity. Training and inference both run efficiently in this form, the authors show. The design also encodes lookup tables, a native use of the old theorem. It stands as a clear, small rival to transformer bulk.
Catalog record
| CREATOR | Sarah Polson, Vadim Sokolov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-00704