Reviews: On the Hardness of Robust Classification

However it has not been made clear how do these impossibility results have any impact from a practical point of view. I would have appreciated if authors have provided some algorithms which achieves robust learnability even for a slightly relaxed setting, e.g. at least a robust learning algorithm for monotone conjugate functions for Thm. Line 200 -- unexpected occurrence of ",", Line 239 "a universal" repeated, Line 310: given -- gave etc. etc. 3. Seemingly incomplete results/ confusing theorem statements: a) Its very confusing as authors claim in the Introduction that: "On the other hand, a more powerful learning algorithm that has access to membership queries can exactly learn monotone conjunctions and as a result can also robustly learn with respect to exact in the ball loss." I would rather have appreciated the result lot more if proved for a more general concept class where sample complexity depends on some complexity measure (e.g. VC dimension) of C_n, etc. 4. Experiments: The paper does not provide any empirical studies, but I understand that this is rather out of the scope of the current results since all the claims are mostly made on non-existence of any robust-learning algorithms, but atleast authors could have proposed some algorithms for the monotone conjugate functions (i.e. for Thm.