Reinforcement Learning with Immediate Rewards and Linear Hypotheses

We consider the design and analysis of algorithms that learn from the consequences of their actions with the goal of maximizing their cumulative reward, when the consequence of a given action is fe...