Experimental Design in Dynamical System Identification: A Bandit-Based Active Learning Approach.
Artémis LlamosiAdel MezineFlorence d'Alché-BucVéronique LetortMichèle SebagPublished in: ECML/PKDD (2) (2014)
Keyphrases
- experimental design
- active learning
- random sampling
- uncertainty sampling
- experimental designs
- machine learning
- sample size
- training examples
- class imbalance
- learning algorithm
- learning strategies
- semi supervised
- learning process
- training set
- supervised learning
- empirical studies
- labeled data
- semi supervised learning
- generalization error
- query by committee
- data sets
- learning systems
- cost sensitive
- feature space
- bandit problems