Exploring Q-learning
Although there are many variants of reinforcement learning, the previous explanation should have given you a good general overview of how most reinforcement models work. It is now time to move deeper into a specific model for reinforcement learning: Q-learning.
Q-learning is a reinforcement learning algorithm that is, so-called, model free. Model-free reinforcement learning algorithms can be seen as pure trial-and-error algorithms: they have no prior notion of the environment, but merely just try out actions and learn whether their actions yield the correct outcome.
Model-based algorithms, on the other hand, use a different theoretical approach. Rather than just learning the outcome based on the actions, they try to understand their environment through some form of a model. Once the agent learns how the environment works, it can take actions that will optimize the reward according to this knowledge.
Although the model-based approach may seem more intuitively...
