Professional Documents
Culture Documents
(a) What is the probability mass function (pmf) for X1 , X2 , . . . , Xn ? Let k denote the number of
1s. Find a function (r, k, n) such that your answer can be written as exp(n (r, k, n)).
(b) What is the probability that exactly k of the Xi are equal to 1?
(c) (*) Minimize the function (r, k, n) over r for a fixed k and n. What is the maximizing value?
Explain why this makes sense in terms of the probability of the sequence. If you fix r and n,
what is the k that maximizes (r, k, n)?
Problem 0.8 Which of the following are valid density functions?
(a) f (x) =
1
1+x2
(b) f (x, y) =
for x R
1
2 2
exp 12
(c) f (x, y) = 3 ex +
(d) f (x) =
1
2
1 2
2x
2 x
3 e
+ y2
for (x, y) R2
for x, y 0
for x {1, 1}
3 2 1
= 2 3 2 .
1 2 3
Find the conditional distribution of (X1 , X2 ) given X3 .
(c) (**) For the same in the previous part, can you find a matrix A such that X = AU where
U N (0, I)?
Problem 0.10 Which of the following functions are convex?
(a) f (z) = max{0, 1 z}
(b) f (z) = log(1 + ez )
(c) f (x, y) = 4x 3y
Problem 0.11 (*) Suppose that a probability distribution over a finite set X = {1, 2, . . . , M } has
the following form. Given a vector of parameters Rd , the probability of x X is given by
P (x; ) = exp > (x) A() ,
where : X Rd is a function that assigns a d-dimensional vector to each x X .
(a) Give a formula for A() in terms of {(x) : x X }.
(b) Calculate A() (the gradient as a function of ).
(c) (**) Calculate 2 A() (the Hessian as a function of ).
(d) (**) Is A() convex? Why or why not?