Suppose when performing linear regression, we cared more about getting some examples correct than on getting other examples correct. To model this we couldweighthe cost of some mistakes more than others.

Our goal is now to find:

N WwIS = arg min w ( 2 ) WERd S (y(2 ) - x (i) I w ) 2 i=1 Find a closed form solution to this problem. (Hint: You can create a matrix Q E RNXN which is a diagonal matrix where Li,i = w(). Now Wwis = arg minweRa (y - Xw)"n(y - Xw).)