144
click to vote
JMLR
2012
13 years 5 months ago
2012
We present a minimax framework for classification that considers stochastic adversarial perturbations to the training data. We show that for binary classification it is equivale...