Revealing Perceptible Backdoors in DNNs, Without the Training Set, via the Maximum Achievable Misclassification Fraction Statistic.
Zhen XiangDavid J. MillerHang WangGeorge KesidisPublished in: MLSP (2020)
Keyphrases
- training set
- classification accuracy
- test set
- training data
- error rate
- nearest neighbor
- satisfiability problem
- cross validation
- hidden structure
- active learning
- supervised learning
- training samples
- decision trees
- data sets
- classification error
- cost sensitive
- test data
- classification algorithm
- misclassification rate
- training examples
- feature space
- training set size
- svm classifier
- class labels
- support vector machine
- sat solvers
- active appearance models
- feature extraction
- learning algorithm