Compact Rule-Based Classifier Learning via Gradient Descent

Javier Fumanal-Idocin, Raquel Fernandez-Peralta, Javier Andreu-Perez · 2025

Read Compact Rule-Based Classifier Learning via Gradient Descent by Javier Fumanal-Idocin, Raquel Fernandez-Peralta, Javier Andreu-Perez (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-2502-01375.txt.

Description

Rules a person can read are gold in high-stakes calls, like loan or safety choices, but teaching a machine to write short rule sets has long fought the jagged nature of rules. Javier Fumanal-Idocin, Raquel Fernandez-Peralta, and Javier Andreu-Perez present a learner in a 2025 paper that trains rule sets with the smooth methods used for neural nets. Their system uses fuzzy logic with meaningful cut points.

One rule alone can settle a case in their design, which keeps the rule book tiny. Across forty data sets the system beat a classic rule learner on accuracy. It matched tree models while using a rule base a tenth the size. Against the best stacked rule systems it kept most of the accuracy with a small fraction of the rules.

Catalog record

CREATORJavier Fumanal-Idocin, Raquel Fernandez-Peralta, Javier Andreu-Perez
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE60 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-01375

Browse the full library · About the Tower of Babel