Uncertainty Scoring
A score estimating sample uncertainty based on prediction entropy, confidence, or Bayesian approximation. High-uncertainty samples typically lie near decision boundaries, and annotating them can effectively correct those boundaries.
Timeline (last 90 days)
Uncertainty scores are often based on predictive entropy, confidence or Bayesian approximation, and high-uncertainty samples often lie near the decision boundary.
This research combines region coverage with uncertainty scoring, achieves leading results in overall accuracy and class balance, and reduces annotation cost.
All Facts (2)
Uncertainty scores are often based on predictive entropy, confidence or Bayesian approximation, and high-uncertainty samples often lie near the decision boundary.
50%UnverifiedThis research combines region coverage with uncertainty scoring, achieves leading results in overall accuracy and class balance, and reduces annotation cost.
50%