The advantages of multiple classes for reducing overfitting from test set reuse.

Vitaly Feldman Roy Frostig Moritz Hardt

Published in: ICML (2019)

Keyphrases

test set
multiple classes
multi class
error rate
training set
training data
test data
cross validation
class distribution
support vector machine
binary classification
multi label classification
classification accuracy
machine learning
decision trees
generalization error
feature selection