Reinforcement Learning (RL) is one of three machine learning techniques. The other two methods are Supervised Learning and Unsupervised Learning. Unlike the other two methods, reinforcement learning is concerned with the evaluation of reactions. Correct reactions are evaluated positively, while incorrect ones are evaluated negatively.
Transferred to image recognition, a positive evaluation is made when reinforcement learning has correctly recognized an object and is able to assign it. For example, if the system correctly recognizes an apple in a fruit bowl, this is evaluated positively. As with the other two methods, reinforcement learning also involves a learning process. In this learning process, an agent determines the environment to which it will respond with a positive or negative action. Depending on how the environment reacts to the response, it is incorporated into future decisions. If the evaluation is positive, reinforcement takes place, reinforcing the decision process; if the evaluation is negative, future decisions will change.