Basic Terminologies of Reinforcement Learning

There are some basic terms you all should know before stepping into Reinforcement learning. Understanding these terms will make you grasp the concept even more efficiently. Follow the diagram given below :

Terminologies of reinforcement learning
  1. Agent: Entity that iteratively interacts with the environment and performs actions. It receives feedback for every action that it performs. For example, the self-driving car in the above diagram is the agent.
  2. Environment: Condition to which the agent interacts iteratively to perform actions. For example, the self-driving car iteratively interacts with the path on which it moves. The path here in this case is the environment.
  3. Action: The task performed by the agent within the environment. It can also describe the change in the state of an agent. For example, The car is moving, is the action.
  4. State: For every action done, a situation is returned to the agent, from the environment. For example, the car needs to stop when it notices a red traffic light.
  5. Policy(π): The strategy panned for the next action by the agent, according to the current action. 
  6. Value: This can be stated as the future rewards an agent would get by being in a particular state. 
  7. QValue: The term is the same as Value, only it accepts current action as an additional parameter.