Contextual bandit is **contextual bandit is** a machine learning framework that addresses sequential decision-making problems where an algorithm selects actions based on observed contextual information to maximize cumulative reward over time. This framework investigates strategies for optimizing action selection in dynamic environments, leveraging contextual features to inform decisions while managing the inherent exploration-exploitation tradeoff. Key characteristics include the use of a policy function mapping context to action choices, operation over discrete time steps, and the objective of minimizing cumulative regret. It is a fundamental concept in online learning and reinforcement learning, widely applied in domains such as personalized recommendations, online advertising, and adaptive experimentation.