Researchers have developed SymboLLM-FE, a novel approach that combines symbolic regression with large language models (LLMs) for automated feature engineering on tabular data. This method aims to overcome the limitations of traditional automated feature engineering, such as poor interpretability and the need for numerous iterations with LLMs, while also mitigating risks of bias and hallucination. By extracting mathematically expressive formulas through symbolic regression and then refining them with LLMs for enhanced interpretability, SymboLLM-FE has demonstrated superior performance over existing methods in six real-world datasets and four Kaggle competitions, achieving high utility with significantly fewer LLM calls. AI
IMPACT This approach could lead to more interpretable and efficient feature engineering for tabular data, potentially improving model performance across various applications.
RANK_REASON The cluster describes a new research paper detailing a novel method for automated feature engineering. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Hugging Face
- Kaggle
- Large language models
- Symbolic regression
- SymboLLM-FE
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
