I'm looking at this true/false question (currently in a ML course) and there's no given solution/explanation for this, I cant really understand what am I supposed to look for when facing this:

"for some family of classifiers $H$, assuming it contains the perfect classifier, we are able to sample a group $S$ from distribution $D$ with enough samples such that the real error over the distribution $D$ is in range $\pm \epsilon$ for all $h\in H$, in a probability of our choosing."

what confuses me the most is the "probablility to our choosing" because up to that part, the statement seems to be true.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.