DAFS: A distribution-aware hierarchical feature selection method for long-tailed classification

Key Points

• Improved accuracy in long-tailed classification, achieving a notable enhancement in performance metrics.
• The method integrates a distribution-aware approach, utilizing hierarchical structures for effective feature selection.
• Assessment focuses on optimizing bias mitigation strategies, leveraging existing datasets across various scenarios.
• Highlights potential for better generalization in machine learning tasks with skewed distributions, expanding applicability.