Machine learning models for power systems show generalization failures

A new paper published on arXiv investigates the generalization failures of machine learning models used for power system protection. The study found that while overall F1 scores remained similar when models were tested on different relays than they were trained on, the false-positive rate significantly increased. This suggests that aggregate metrics can mask critical location-specific changes in alarm behavior, emphasizing the need to evaluate different error types separately. AI

IMPACT Highlights the need for careful evaluation of ML models in critical infrastructure to avoid masked failure modes.

RANK_REASON The cluster contains a research paper published on arXiv detailing findings about machine learning model generalization. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Bus 2 (Pauly-Wissowa)
• Bus 300 affair
• F1 scores
• False negative rates in Drosophila cell-based RNAi screens: a case study
• false-positive rate (FPR)
• Hypothetical protein CELE_R03H4.6
• machine learning
• multilayer perceptron
• Nithin Pradeep Nadayil Kizhakkethil
• power-system protection

• paper
• other

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