Answer goes here Part (6) MAP estimation vs. loss minimization Recall that for linear regression we are minimizing a loss function L(w) = 21-(w?x; 3:)? and that this corresponds to maximizing the log-likelihood logp(y|X, w). Answer the following two questions: 1. Show that minimizing the regularized loss function Lx(w) = ||w v|l2+ (w?x; 3:)? corresponds to finding the maximizer of the posterior distribution (maxiumum a posteriori or MAP estimator) w* = argmax p(w|X,y) WERP when the prior distribution for w is a Gaussian N(v, yI), where I RPXP is the identity matrix and y> 0. 2. What is the relation between 1, 02, and y?? In particular, what does mean for the loss function if the prior distribution is very concentrated around the mean v (i.e. y is close to 0)? Answer goes here Extemally added files can be added to Git Visu Files Away's Add Don't Asl: Again Answer goes here Part (6) MAP estimation vs. loss minimization Recall that for linear regression we are minimizing a loss function L(w) = 21-(w?x; 3:)? and that this corresponds to maximizing the log-likelihood logp(y|X, w). Answer the following two questions: 1. Show that minimizing the regularized loss function Lx(w) = ||w v|l2+ (w?x; 3:)? corresponds to finding the maximizer of the posterior distribution (maxiumum a posteriori or MAP estimator) w* = argmax p(w|X,y) WERP when the prior distribution for w is a Gaussian N(v, yI), where I RPXP is the identity matrix and y> 0. 2. What is the relation between 1, 02, and y?? In particular, what does mean for the loss function if the prior distribution is very concentrated around the mean v (i.e. y is close to 0)? Answer goes here Extemally added files can be added to Git Visu Files Away's Add Don't Asl: Again