Decomposable Neuro Symbolic Regression
Giorgio Morales [email protected], John W. Sheppard [email protected] · 2025
Read Decomposable Neuro Symbolic Regression by Giorgio Morales [email protected], John W. Sheppard [email protected] (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-2511-04124.txt.
Introduction
Giorgio Morales and John W Sheppard present decomposable neural symbolic regression, a method that explains a trained opaque regression model by turning it into readable math expressions. Plain symbolic regression often chases prediction error and returns bloated or wrong formulas.
A multi-set transformer drafts one-variable skeletons showing how each variable moves the opaque model's response. Genetic search selects strong skeletons, genetic programming merges them while keeping structure, and coefficients are tuned last. Tests under set, varying noise check whether the recovered formulas track the true links rather than merely fitting points.
Catalog record
| CREATOR | Giorgio Morales [email protected], John W. Sheppard [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 175 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-04124