Say you have a bandit problem where the only feedback is

a) the number of times you pulled arm A

b) the number of times you pulled arm B

c) the sum total reward associated with A and B

on a daily basis. What strategy would you use to obtain the best policy here? I was thinking of just training a linear model where A and B are the inputs and C is the output - this weights would tell us the rate of reward for each arm.

  • $\begingroup$ If the bandit arms are deterministic, can’t you just figure out the arm values after two days of data? E.g. two equations for two unknowns in terms of total reward. $\endgroup$ – Alex R. Jul 16 '19 at 23:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.