2-13

• 0:06:48: Q1. What happen if we initialize all weights to 0? ANS: Output equal to 0 and no learning can be done also as gradient of L with respect to weight is proportional to input, which is 0.

• 0:31:04: Q2 why the last equality is true? ANS: because wi are independent, hence wi x_i are independent and consequently uncorrelated

• 0:41:24: Q3 Do you know the formula Var(X)=E[X^2]-E[X]^2? ANS:

• 1:17:24: Q4. what do you expect the variance of hat{x}? ANS: 1