NarmedBanditLearner

This C++ example code snippet shows how to implement a Learner class that has to optimize its reward from an N-armed bandit. As one would expect, the Learner class knows nothing about the N-Armed Bandit class' internals.

Give a Learner a set of actions, with every action having a certain expected reward. Actions taken are evaluated by adjusting the expected rewards. The chance an action is equal to its expected reward of the sum of expected rewards.

The Learner class has an internal Strategy Design Pattern, which enables the user to select a different Learning Strategy at run-time. In the example main, this Strategy is set randomly.