Which learning method uses feedback to optimize actions?

Reinforcement learning is a learning method that focuses on making decisions through trial and error, where an agent learns to perform actions in an environment to maximize a cumulative reward. This method emphasizes the concept of feedback, as the agent takes actions and receives evaluations in the form of rewards or penalties based on those actions.

The feedback allows the agent to understand which actions are beneficial and which are not, guiding it to optimize its decision-making process over time. This learning paradigm mimics how humans and animals learn from their environment, where positive outcomes reinforce specific behaviors, while negative outcomes discourage others. The agent uses this feedback to develop a policy that maximizes its total expected reward.

In contrast, other learning methods like unsupervised learning focus on finding patterns in data without any feedback on actions, while supervised learning relies on labeled data for training without a self-driven optimization process based on an agent's actions in an environment. Deep learning is a subset of machine learning that may be used within supervised or unsupervised frameworks, but it does not inherently involve the reinforcement learning feedback loop. Thus, reinforcement learning distinctly stands out for its reliance on feedback to improve and optimize actions.