A mathematical framework for modelling sequential decision-making where outcomes are partly random and partly under the control of a decision maker, comprising states, actions, transition probabilities, and a reward function, solved by computing a policy that maximises expected cumulative reward.
Semantic Classification
Content
- A Markov decision process extends a Markov chain with actions and rewards, modelling an agent that chooses actions to maximise cumulative reward over time. Its solution is a policy mapping states to actions, often computed via dynamic programming.
- It provides the formal foundation for reinforcement learning, where the transition and reward functions are typically unknown and must be learned through interaction.