Information flow structure. Reinforcement learning is a cause-and-effect learning using the interaction with the environment. Learning by interaction is a fundamental idea in all learning strategies. The objective is how to convert a situation into action and to maximize a numerical reward signal. Trial-and-error search and delayed reward are two important features of reinforcement learning. At high level, reinforcement learning has the following main elements: an agent, an environment, a policy, a reward signal, a value function and optionally a model of the environment. Commented [K(43]: selects An agent is both learner and decision maker; it must sense the state of its environment at some level and take actions with a goal (objective) to achieve the desired impact on the state. The agent must be able to learn from its own experience and make a trade-off between exploration and exploitation. The thing an agent interact with is called environment. An agent selects actions and the environment responds to it and presents its new situation to the agent. The agent cannot influence/impact the rules of the environment.
