On this page of Shai Shalev-Shwartz's ppt, I find it is hard for me to understand the theorem and exersise part.

In the theorem part, I know that the first part is always true, since if a sample $x$ in the data set is labled with 1, we can use this $x-\epsilon$ as the $\theta$, and the trivial condition is that the sequence only have one example, that is $x$. But I can't understand and prove the later part. I mean, what is the meaning of "always?"

In the exercise part, I can't see why we can't learn the class of axis-aligned rectangles. Additionally, does the exercise here means that the function $f$ we learned should be the real predicator? That is,all points in $\mathbb{R}$ should be consistent with $f$?

enter image description here



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.