ResearchResearch paperLarge Language Models · Training & Scaling · Efficiency & Inference1 source · Oct 17, 2026

From Overparameterised Expressions to Parsimonious Models: Fisher-Geometric Pruning for Symbolic Regression

We introduce Fisher-Geometric pruning (FGP), a post-hoc model-reduction framework that acts on fitted models and can complement these techniques.

Genetic programming symbolic regression (GPSR) often returns expressions that are larger than necessary, with dormant branches and finely tuned algebraic configurations that fit sampling noise.
Existing complexity-control strategies, such as multi-objective size-accuracy search or information criteria embedded in the fitness function, act on syntactic representations during model search.
A taxonomy of parametric directions separates local sensitivity from proximity to reducible boundaries, enabling the treatment of unidentifiable directions and reductions through zero constraints, asymptotic limits, and signed parameter identifications.
We evaluate FGP on controlled diagnostic problems and Pareto fronts generated by GPSR.

[1]From Overparameterised Expressions to Parsimonious Models: Fisher-Geometric Pruning for Symbolic Regression
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 17, 03:46 PM
We introduce Fisher-Geometric pruning (FGP), a post-hoc model-reduction framework that acts on fitted models and can complement these techniques.
Genetic programming symbolic regression (GPSR) often returns expressions that are larger than necessary, with dormant branches and finely tuned algebraic configurations that fit sampling noise.

Extractive summary: sentences quoted from the sources.

Before this