From the course: Reinforcement Learning Foundations

Terms in reinforcement learning

From the course: Reinforcement Learning Foundations

Terms in reinforcement learning

- [Instructor] Before we get into the technical details, you first need to understand some very important terms that are common in reinforcement learning. This will give you a better foundation for the rest of the course. First, let's talk about the agent. This is the main property of a reinforcement learning problem. For example, it could be a robot learning to walk, or an agent learning to drive. The agent is the artificially intelligence corrector that does the learning by exploring the environments and observing, just the way we do sometimes. The next term is the state. This is the position the agent is at a given period. It changes when the agent moves. Then the environment is the environment, literally, that the agent observes to learn. Different positions in the environment represent the state. This model of the environment is sometimes known by the agent, and in other cases, unknown. Next is the action. This is whatever step or act the agent decides to take after observing the environment. If this action taken by the agents is correct, it gets a positive reward. Now, reward, in this case, is the score an agent gets for taking an action. Correct actions lead to positive rewards, while wrong actions lead to negative rewards. When an agent fails and gets a negative reward, it learns from it and changes its actions with the aim of choosing the right action. Therefore, the agent tries to tilt from actions that lead to negative rewards to those that lead to positive rewards. Before selecting an action, the agent uses a strategy or method to decide on what action to take in a particular state. They're usually different strategies, and these strategists in reinforcement learning are known us policies. Finally, when the agents explores an environment, it has a mission; it has something it wants to learn and that's its goal. That's enough for now. You'll learn about some other terms as we progress through the course.

Contents