2. [15 points] Suppme that the VC dimension of our hypothesis set \"H is dv = 3 (e.g., linear classiers in R2] and that we have an algorithm for selecting some if E 'H based on a training sample of size n {i.e., we have 1: example inputrontpnt pairs to train on]. (:1) Using the generalization bound given in class, give an upper bound (which depends on RAF?\" on RU?) that holds with probability at least 0.95 in the case where n = lUU. Repeat for n = 1,000 and n = l,. (b) Again using the generalization bound given in class, how large does ' need to he to obtain a generalization bound of the form ea?) 5 111m") + 0.1 that holds with probability at least 0.95? How does this compare to the \"rule of thumb\" given in class