Evading classifiers in discrete domains with provable optimality guarantees.
Bogdan KulynychJamie HayesNikita SamarinCarmela TroncosoPublished in: CoRR (2018)
Keyphrases
- continuous domains
- training data
- support vector
- application domains
- optimal solution
- multiple classifiers
- linear classifiers
- training samples
- classification systems
- feature selection
- real world
- machine learning algorithms
- supervised classification
- discrete geometry
- approximation guarantees
- classification algorithm
- svm classifier
- test set
- training examples
- naive bayes
- decision trees
- logistic regression
- classification method
- classification models
- state space
- np hard
- discrete space
- training set