Cost-Sensitive Self-Training for Optimizing Non-Decomposable Metrics.
Harsh RangwaniShrinivas RamasubramanianSho TakemoriKato TakashiYuhei UmedaVenkatesh Babu RadhakrishnanPublished in: CoRR (2023)
Keyphrases
- cost sensitive
- directly optimize
- multi class
- uncertainty sampling
- cost sensitive learning
- cost sensitive classification
- misclassification costs
- binary classification
- active learning
- class distribution
- naive bayes
- class imbalance
- support vector machine
- confidence weighted
- probability estimates
- machine learning
- data sets
- multi label
- image segmentation
- feature selection