CART regression trees can achieve spatial adaptation with MID stopping rule, study finds

A new paper published on arXiv details the statistical role of stopping rules in CART regression trees. Researchers proved that the minimum impurity decrease (MID) stopping rule, when combined with an appropriate threshold, can achieve minimax rates for spatial adaptation under heterogeneous and anisotropic smoothness conditions. However, the study also found that the widely used minimum leaf size stopping rule cannot achieve this spatial adaptation. AI

IMPACT Provides theoretical grounding for the effectiveness of specific stopping rules in regression tree algorithms, potentially influencing future model development.

RANK_REASON Academic paper on statistical methods for regression trees. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv stat.ML →

• arXiv
• Bayesian Methods
• empirical risk minimization
• Hugging Face
• minimum impurity decrease
• minimum leaf size

• paper
• other

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