//
sign in
Post
by @danabra.mov
PostEmbed
by @danabra.mov
Record
by @jimpick.com
Record
by @atsui.org
+ new component
Post
That is, in certain cases there is no way to post-process a multiaccurate predictor to get a weak learner, even assuming the best hypothesis in C has high correlation. However, when we also require p to be calibrated, we recover not just weak, but strong agnostic learning.