Hands-On Unsupervised Learning with Python

Reinforcement learning algorithms

A reinforcement learning scenario can be considered as a supervised one where the hidden teacher provides only approximate feedback after every decision of the model. More formally, reinforcement learning is characterized by continuous interaction between an agent and an environment. The former is responsible for making decisions (actions), finalized to increase its return, while the latter provides feedback to every action. Feedback is generally considered as a reward, whose value can be either positive (the action has been successful) or negative (the action shouldn't be repeated). As the agent analyzes different configurations of the environment (states), every reward must be considered as bound to the ...