Suppose y = 10, 4, 8 when x is 1,1,2. Our model uses parameters theta=1,1 and when we run it using x=1,1,2 it predicts y = 9,5,8.

We then back calculate dloss/dtheta = 2,-1.

a) In what direction should we move theta to try to reduce loss?

b) Assuming loss was defined as the sum of the squared errors, approximately what would be the sum of the squared errors in our next prediction if we change theta to 1.001,1.001

Answer rating: 100% (QA)

a Since the partial derivatives of the loss function with respect to the parameters theta are dlossd... View the full answer