Consider the complete log-likelihood of the Gaussian mixture model: K K K log P(x, z) = > >zik log TK- SEE zik log(2no)) - 7 (Ti - HK) 2. i-lk-1 ilk- 1 (a) (3 pts.) Suppose z is observed, compute the maximum likelihood estimator for / and a (i.e., maximize log P(, =) over / and o'), where u and o' are K-dimensional vectors. (b) (2 pts.) Given the parameters n, p, and o', compute P(zik = 1/x;)