New metric Feature Instability (FI) enhances model stability analysis

Researchers have introduced a new metric called feature instability (FI) to analyze algorithmic stability, focusing on how sensitive models are to the removal of a single feature. This metric, analogous to instance instability (II), provides complementary information for understanding generalization. Experiments with linear models and random forests demonstrate that feature bagging, an ensemble technique using random feature subsets, significantly improves stability compared to methods without bagging, especially with more aggressive subsampling. AI

IMPACT Introduces a new metric for analyzing model stability, potentially leading to more robust machine learning algorithms.

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

Read on arXiv stat.ML →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Feature Instability
• Gotit.pub
• Hugging Face
• Instance Instability
• random forest
• ScienceCast

• paper
• other

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