Cost-Sensitive Self-Training for Optimizing Non-Decomposable Metrics.
Harsh RangwaniShrinivas RamasubramanianSho TakemoriKato TakashiYuhei UmedaVenkatesh Babu R.Published in: NeurIPS (2022)
Keyphrases
- cost sensitive
- directly optimize
- misclassification costs
- multi class
- cost sensitive learning
- uncertainty sampling
- cost sensitive classification
- class distribution
- binary classification
- active learning
- probability estimates
- naive bayes
- confidence weighted
- class imbalance
- support vector machine
- neural network
- boosting algorithms
- base classifiers