A solvable teacher-student perceptron model predicts that the optimal fraction of anomaly examples in training is generally away from 50%, with a sharp crossover as training noise increases.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Class Imbalance in Anomaly Detection: Learning from an Exactly Solvable Model
A solvable teacher-student perceptron model predicts that the optimal fraction of anomaly examples in training is generally away from 50%, with a sharp crossover as training noise increases.