LoBoost method enhances uncertainty quantification for gradient-boosted trees

Researchers have introduced LoBoost, a novel method for local conformal prediction designed to enhance uncertainty quantification for gradient-boosted decision trees. This approach leverages the internal structure of the fitted ensemble, specifically its leaf partitions, to estimate residual quantiles locally. Unlike standard conformal prediction, LoBoost adapts better to heteroscedasticity without requiring additional data splits, retraining, or auxiliary models. Experiments demonstrate that LoBoost achieves competitive interval quality and stable behavior across various local calibration settings. AI

IMPACT Improves uncertainty quantification for tabular data models, potentially leading to more reliable predictions in critical applications.

RANK_REASON The cluster contains an academic paper detailing a new method for machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• arXiv
• Conformal prediction
• DagsHub
• Hugging Face
• LoBoost
• Victor Coscrato

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →