Reinforcement learning

Definition

A method for optimizing long-term cumulative return by interacting with an environment, receiving rewards and improving an action policy.

Reinforcement learning addresses sequential problems in which actions affect subsequent states and returns. Q-learning is a representative approach to value learning, with its 1992 paper providing corresponding theoretical results; this is not the year reinforcement learning originated.

Reward definitions, exploration methods and assumptions about the environment determine learning outcomes. Convergence results under specific conditions cannot be extended directly to arbitrary deep models or open environments.

Principles and impact

Problems solved

Learns sequential decision-making from delayed feedback without requiring a correct-action label at every step.

Principles

Establishes mechanisms for updating values or policies based on states, actions and rewards.

Improvements

Not recorded

Significance

Connects machine learning with goal-directed sequential decision-making.

History and applications

People

Christopher J. C. H. Watkins、Peter Dayan

Organizations

Not recorded

Events

Not recorded

Applications

Sequential decision-making、Control

Limitations

Reward specifications may diverge from the real objective, while exploration costs and distribution shifts constrain practical use.

Upstream and downstream relations

Upstream

←Machine learningspecializesReinforcement learning learns to make decisions from rewards received through interaction with an environment and is the branch of machine learning concerned with sequential decision-making.

Sources: Q-learning

Downstream

→Deep reinforcement learninghas specialized formDeep reinforcement learning uses deep function approximation for value or policy learning and is a class of reinforcement learning methods.

Sources: Human-level Control through Deep Reinforcement Learning

→Reinforcement learning from human feedbackdrawn on byRLHF borrows the optimization framework of reinforcement learning, replacing entirely hand-specified objective signals with rewards learned from human preferences.

Sources: Deep Reinforcement Learning from Human Preferences

Sources for this entry and its relations

Human-level Control through Deep Reinforcement Learning

Authors / institution: Volodymyr Mnih et al.

Published: 2015.

Original text / resources: Human-level Control through Deep Reinforcement Learning

Scope of evidence: Supports the use of deep networks to approximate action values and the combination of pixel inputs with reinforcement learning. Results in specific games do not establish general capability in real-world environments.

Accessed: 2026-10-13.

Deep Reinforcement Learning from Human Preferences

Authors / institution: Paul F. Christiano et al.

Published: 2017.

Original text / resources: Deep Reinforcement Learning from Human Preferences

Scope of evidence: Supports learning a reward function from human comparisons of trajectory segments, then using it for reinforcement learning.

Accessed: 2026-10-13.

Q-learning

Authors / institution: Christopher J. C. H. Watkins, Peter Dayan.

Publication date: 1992.

Original text / materials: Q-learning

Scope of evidence: Supports action-value learning from rewards received through interaction. Q-learning is used here only as a representative reference point for reinforcement learning, not to assert that reinforcement learning originated in 1992.

Accessed on: 2026-10-13.

Definition

Principles and impact

Problems solved

Principles

Improvements

Significance

History and applications

People

Organizations

Events

Applications

Limitations

Upstream and downstream relations

Upstream

Downstream

Explore this technology's connections