Improved cross-validation for classifiers that make algorithmic choices to minimise runtime without compromising output correctness.
Dorian FlorescuMatthew EnglandPublished in: CoRR (2019)
Keyphrases
- cross validation
- support vector
- training set
- unseen data
- model selection
- nearest neighbor classifiers
- variable selection
- binary classifiers
- roc analysis
- hyperparameters
- cross validated
- training data
- generalization error
- error estimates
- test set
- regularization parameter
- meta learning
- decision trees
- feature selection
- svm classifier
- grid search
- input variables
- class labels
- machine learning algorithms
- kernel function
- training samples
- data sets
- rbf kernel
- face recognition
- feature space
- small sample
- k nearest neighbour
- nearest neighbor
- ls svm
- feature set
- em algorithm
- test data
- logistic regression