Consider two-bandit problem with the following reward distributions: ~ R(a) Uniform[0 1.4] R(a2) N(u=0.5, = 1)...

 

Transcribed Image Text:

Consider two-bandit problem with the following reward distributions: ~ R(a) Uniform[0 1.4] R(a2) N(u=0.5, = 1) ~ a) Compute the optimal Q*(a), Q*(a) and *. b) Consider the reward distributions are unknown. Use the learning rate = 0.5 to estimate Q(a), Q(a) and given the following: k=1 k=2 k=3 k=4 k=5 Action a a a a Reward 1 0.5 0 1.25 1.35 c) Repeat part b for optimistic initial value Given Q(a) = Q(a) = 5. Consider two-bandit problem with the following reward distributions: ~ R(a) Uniform[0 1.4] R(a2) N(u=0.5, = 1) ~ a) Compute the optimal Q*(a), Q*(a) and *. b) Consider the reward distributions are unknown. Use the learning rate = 0.5 to estimate Q(a), Q(a) and given the following: k=1 k=2 k=3 k=4 k=5 Action a a a a Reward 1 0.5 0 1.25 1.35 c) Repeat part b for optimistic initial value Given Q(a) = Q(a) = 5.