What is Reinforcement Learning?
Reinforcement Learning is a field of machine learning in which an agent chooses from a set of actions in an environment with the goal of maximizing future rewards over time. It is used in robotics, games, personalized recommendations and many other areas.
Key elements in reinforcement learning:
Agent: it is the neural network that is trained to perform the actions.
Environment: it is the world where the agent performs its actions.
Action: a move that the agent can perform in the environment. For example: walk forward, walk backwards, jump.
Observation: it is the feedback that the agent receives from the environment.
Reward: it is the evaluation that the agent gets for the action that it performed.
One example that can help understand how reinforcement learning works is a toddler that is learning how to walk. When walking, toddlers adjust their steps based on the outcome they get. Perhaps taking a smaller step will help them have more stability so they will be doing more smaller steps to avoid falling all the time. In the same way, agents use reinforcement learning algorithms to determine the best actions that they should take in order to get the most rewards from the environment.
References