To demonstrate the basic Q-learning algorithm, let's consider a simple problem. Imagine that our agent (player) lives in a grid world. One day, she was trapped in a weird maze, as shown in the following diagram:
The maze contains six rooms. Our agent appears in Room 1, while she has no knowledge about the maze, that is, she doesn't know Room 6 has the sweetheart that is able to send her back home, or that Room 4 has a lightning bolt that strikes her. Therefore, she has to explore the maze carefully to escape as soon as possible. So, how do we make our lovely agent learn from experience?
Fortunately, ...
