Reinforcement learning Introduction
Agents & rewards
Learn actions through interaction and feedback.
Before this
The intuition
An agent selects an action, observes what happens, and receives a reward. A policy specifies how the agent chooses actions. The objective usually concerns rewards over time.
A concrete example
A maze agent may receive −1 per step and +10 for reaching an exit. A shorter successful route earns a larger total reward under this rule.
What to remember
Reward design matters. Maximizing a poorly chosen reward may produce unintended behavior.
Check your understanding
Why might the best action have no immediate reward?
Reveal explanation
It can lead to a state with better future rewards.