Chapter 15. Deep Reinforcement Learning

Reinforcement learning is a form of learning in which a software agent observes the environment and takes actions so as to maximize its rewards from the environment, as depicted in the following diagram: 

This metaphor can be used to represent real-life situations such as the following:

• A stock trading agent observes the trade information, news, analysis, and other form information, and takes actions to buy or sell trades so as to maximize the reward in the form of short-term profit or long-term profit.
• An insurance agent observes the information about the customer and then takes action to define the amount of insurance premium, so as to maximize the profit and minimize the risk.
• A humanoid robot observes the environment and then takes action, such as walking, running, or picking up objects, so as to maximize the reward in terms of the goal achieved.

Reinforcement learning has been successfully applied to many applications such as advertising optimization...