Halfspace depth equals the minimal 0-1 classification risk of a linear classifier on Q plus a single negative point, and replacing the loss or classifier yields new 'loss depths' that perform competitively in anomaly detection.
Random rotation ensembles
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Data Depth as a Risk
Halfspace depth equals the minimal 0-1 classification risk of a linear classifier on Q plus a single negative point, and replacing the loss or classifier yields new 'loss depths' that perform competitively in anomaly detection.