84117275be999ff55a987b9381e01f96-Reviews.html

Assigned_Reviewer_2 This response refers to the initial review. The new review differs substantially from the original one and corrects significant misunderstandings. The following claim by the reviewer demonstrates the misunderstanding: "In general, I am not convinced that this new model is especially fundamental, novel, or helpful. While the authors claim this to be a generalization of the standard SQ model, the opposite can also be claimed [...] Specifically, to implement an'active statistical' query, f,g, one can simply query a standard SQ oracle with the function h(x,y) g(x,y)*f(x), and then divide by the measure induced by f, namely the response to the SQ oracle f(x)." What the reviewer has overlooked however is that such classic SQ implementation requires label-dependent queries of much higher tolerance.