Provable Robustness of Adversarial Training
for Learning Halfspaces with Noise
Difan Zou * 1 Spencer Frei * 2 Quanquan Gu 1
Abstract To formalize the above comment, let us define the ro-
We analyze the properties of adversarial train- bust error of a classifier. Let D be a distribution over
ing for learning adversarially robust halfspaces (x, y) ∈ Rd × {±1}, and let f : Rd → {±1} be a hy-
in the presence of agno ...


雷达卡




京公网安备 11010802022788号







