Adversarial vs behavioural-based defensive AI with joint, continual and active learning: automated evaluation of robustness to deception, poisoning and concept drift.
Alexandre DeyMarc VelayJean-Philippe FauvelleSylvain NaversPublished in: CoRR (2020)
Keyphrases
- concept drift
- active learning
- data streams
- streaming data
- class imbalance
- drift detection
- drifting concepts
- evolving data streams
- data stream classification
- incremental clustering
- change detection
- machine learning
- data distribution
- classification algorithm
- non stationary
- data stream mining
- semi supervised
- batch learning
- sliding window
- semi supervised learning
- pairwise
- high quality
- feature extraction
- learning algorithm