New framework enhances OOD detection by optimizing model, data, and representation learning

A new research paper introduces a framework to improve out-of-distribution (OOD) detection in machine learning models. The proposed method optimizes three key factors: model reminder, data sampling, and representation learning. It includes Self-Knowledge Distillation (SKD) to maintain classification accuracy, Semi-hard Outlier Sampling (SOS) for efficient outlier detection with minimal data, and Outlier-aware Supervised Contrastive Learning (OSCL) to enhance the separation between in-distribution and OOD data. This approach demonstrates improved performance and accuracy across various benchmarks, especially in long-tailed scenarios. AI

IMPACT Improves robustness of AI models in real-world scenarios by enhancing out-of-distribution detection capabilities.

RANK_REASON Academic paper detailing a new method for OOD detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• HyunJun Choi
• Outlier-aware Supervised Contrastive Learning
• Self-Knowledge Distillation
• Semi-hard Outlier Sampling

• paper
• model release

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