The Adversarial Consistency of Surrogate Risks for Binary Classification

Neural Information Processing Systems 

We study the consistency of surrogate risks for robust binary classification.