1 paper
Changyu Liu, Yuling Jiao, Junhui Wang +1
We propose a general approach to evaluating the performance of robust estimators based on adversarial losses under misspecified models. We first show that adversarial risk is equiv…