A Linear Response Bandit Problem