Symbolic Regression for Shared Expressions: Introducing Partial Parameter Sharing
Researchers have developed a new method for symbolic regression that allows for partial parameter sharing across multiple categorical variables. This approach enables the discovery of single expressions that can describe related phenomena while differentiating between universal effects, category-specific trends, and interactions. The method was tested on synthetic data and an astrophysics dataset, demonstrating its ability to achieve similar fit quality with fewer parameters and extract additional information. AI
IMPACT Introduces a novel technique for symbolic regression, potentially enhancing interpretability and efficiency in scientific discovery.