Classification Under Misspecification: Halfspaces, Generalized Linear Models, and Evolvability
NIPS 2020, 2020.
Our results are perhaps most closely related to the notion of equal opportunity, where our experiments show that many off-the-shelf algorithms achieve high false negative rate on certain demographic groups when noise is added to the rest of the data
In this paper we revisit some classic problems on classification under misspecification. In particular, we study the problem of learning halfspaces under Massart noise with rate $\eta$. In a recent work, Diakonikolas, Goulekakis, and Tzamos resolved a long-standing problem by giving the first efficient algorithm for learning to accuracy $...More
PPT (Upload PPT)