2. [Gradient descent] (5 pts) You are performing gradient descent with momentum to train a model with two learnable parameters w = w1 w2 . Suppose that the loss is L(w) = (w1 - 1)2 + - (W2 - W1 ) 2 . Recall that the update rule is Pt+1 + H . pt - QV wL(wt) Wt+1 + Wt+ Pt+1. Suppose we use hyperparameters o = 0.1 and u = 0.9, and that we initialize p1,1 = P1,2 = 0 and w1,1 = W1,2 =0. What is the value w3 (i.e., take two gradient steps)