Challenges of Reliable, Realistic and Comparable Active Learning Evaluation.
Daniel KottkeAdrian CalmaDenis HuseljicGeorg KremplBernhard SickPublished in: IAL@PKDD/ECML (2017)
Keyphrases
- active learning
- evaluation model
- real life
- random sampling
- cost effective
- real world
- high quality
- open issues
- key issues
- imbalanced data classification
- neural network
- selective sampling
- gold standard
- learning strategies
- unlabeled data
- supervised learning
- semi supervised
- mobile devices
- training set
- reinforcement learning
- training data