h,b(x)=Tx+b with xR784,R784 and bR. This time we will use the logistic loss instead of the squared loss. Instead of coding everything from scratch, we will also use the package scikit learn and study the effects of 1 regularization. You may want to check that you have a version of the package up to date (0.24.1). 26. Recall the definition of the logistic loss between target y and a prediction h,b(x) as a function of the margin m=yh,b(x). Show that given that we chose the convention yi{1,1}, our objective function over the training data {xi,yi}i=1m can be re-written as L()=2m1i=1m(1+yi)log(1+eh,b(xi))+(1yi)log(1+eh,b(xi)). 27. What will become the loss function if we regularize the coefficients of with an 1 penalty using a regularization parameter