[Solved] IE613 Assignment2

Question 1 Suppose that X is subgaussian and X₁ and X₂ are independent and ₁ and ₂subgaussian respectively, then:

1. E[X]=0 and Var[X] ².
2. cX is |c|subgaussian for all c R.

Question 2 Suppose that X is zero-mean and X [a,b] almost surely for constants a < b.

1. Show that X is (b a)/2
2. Using Cramer-Chernoff method shows that if X₁,X₂,,X_n are independent and X_t [a_t,b_t] almost surely with a_t < b_t for all t, then prove

Question 3 [Expectation of maximum] Let X₁,,X_n be a sequence of -subgaussian random variables (possibly dependent) and Z = max_t[n] X_t. Prove that

1.

2.for any (0,1).

Question 4 [Bernsteins inequality] Let be a sequence of independent random vari-

n

ables with X_t E[X_t] b almost surely and S = ^P(X_t E[X_t]) and v = ^P V [X_t].

t=1 t=1

1. Show that is increasing.
2. Let X be a random variable with E[X] = 0 and X b almost surely. Show that E[exp(X)] 1 + g(b)V [X].
3. Prove that for all 0. Prove that this is the best possible approximation in the sense that the 2 in the denominator cannot be increased.

2-1

2-2 Homework 2: March 19

4. Let and and prove that
5. Use the previous result to show that

Question 5 Show that

implies the regret of an optimally tuned Explore-then-Commit (ETC) algorithm for subgaussian 2armed

bandits with means ₁,₂ R and = |_{1 2}|, satisfies R_T + C T where C > 0 is a universal constant.

Question 6Fix (0,1). Modify the ETC algorithm to depend on and prove a bound on the pseudo-regret of ETC algorithm that holds with probability 1 where A_t is the arm chosen in the round t.

Hint: Choose m appropriately in the regret upper bound of ETC algorithm which is proved in the class.

Question 7 Fix (0,1). Prove a bound on the random regret of ETC algorithm that holds with probability 1 . Compare this to the bound derived for the pseudo-regret in the question 5. What can you conclude?

Question 8 Assume the rewards are 1subgaussian and there are k 2 arms. The greedy algorithm depends on a sequence of parameters. First it chooses each arm once and subsequently chooses A_t = argmax_i(t 1) with probability and otherwise chooses an arm uniformly at random.

i

1. Prove that if, then.
2. Let _min = min{_i : _i > 0} where _i = ^? _i, and where C > 0 is a sufficiently large universal constant. Prove that there exists a universal C⁰ > 0 such that

.

Question 9 Fix a 1subgaussian karmed bandit environment and a horizon T. Consider the version of UCB that works in phases of exponentially increasing length of 1,2,4,. In each phase, the algorithm uses the action that would have been chosen by UCB at the beginning of the phase.

Homwork 2: March 19 2-3

1. State and prove a bound on the regret for this version of UCB.
2. How would the result change if the l^th phase had a length of d^le with > 1?