4. We showed that the Support Vector Machine (SVM) can be viewed as minimizing hinge loss: minw,bi=1NLH(y,t)+21w22 where hinge loss is defined as: LH(y,t)=max(0,1ty) Here t{1,1} is the label, y=wTx+b is the predicted score. (a) TRUE or FALSE: if the total hinge loss is zero, then every training example must be classified correctly. Justify your answer. (b) Suppose we replace the hinge loss with the following: L(y,t)=max(0,ty)