. 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 occurs when reinforcement learning has correctly recognized an object
and can assign it to it.For example, if the system correctly recognizes an apple in a fruit bowl, then 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, a reinforcement takes place, whereby the decision process is strengthened; if the evaluation is negative, future decisions will change. The algorithms for reinforcement learning take place in the Markov process, taking into account the environment and actions.