Feature bagging enhances algorithmic stability, study finds

Researchers have analyzed feature bagging, an ensemble method that trains base learners on randomly subsampled feature subsets, focusing on its impact on algorithmic stability. They introduced "feature instability" (FI) as a measure analogous to "instance instability" (II), finding that lower FI and II values indicate greater stability. Experiments in both parametric and model-free settings demonstrated that feature bagging enhances stability compared to non-bagged approaches, with more aggressive subsampling yielding larger improvements. The study also indicated that a moderate number of bagging rounds can achieve stability levels close to infinite bagging. AI

IMPACT This research provides theoretical guarantees for feature bagging, potentially improving the robustness and generalization of machine learning models.

RANK_REASON The cluster contains an academic paper detailing a new method and analysis in machine learning.

Read on Hugging Face Daily Papers →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Feature bagging for outlier detection
• Feature Instability
• Gotit.pub
• Hugging Face
• Instance Instability
• random forest
• ScienceCast
• algorithmic stability

• paper
• other

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →