Importance-weighted label prediction for active learning with noisy annotations.
Liyue ZhaoGita SukthankarRahul SukthankarPublished in: ICPR (2012)
Keyphrases
- active learning
- label noise
- prediction accuracy
- active learning framework
- random sampling
- prediction algorithm
- learning strategies
- labeling effort
- prediction error
- prediction model
- relative importance
- semi supervised
- multiple annotators
- semantic labels
- experimental design
- image labeling
- meta information
- weighted graph
- annotation tool
- training set
- metadata
- neural network
- unlabeled instances
- active learner
- pool based active learning
- selective sampling
- noise free
- cost sensitive
- image annotation
- missing data
- training examples
- semi supervised learning
- supervised learning
- learning algorithm
- machine learning