Reinforcement Learning

Machine learning where an agent takes actions in an environment to maximize cumulative reward — the engine behind AlphaZero and its descendants.

Specifications & Metrics

type

Learning Methods & Systems

symbol

RL

Core Principles & Analysis

• A machine learning paradigm in which an agent takes actions in an environment to maximize cumulative reward.
• The problem is classically formulated as a Markov decision process, with a policy mapping states to actions.
• It suits sequential decision-making problems amenable to planning and combinatorial search spaces with high-dimensional inputs.

Sources

• Wikipedia: Reinforcement learning