Reinforcement Learning (RL)

Reinforcement learning is a type of machine learning where an agent learns by trial and error, taking actions in an environment and receiving rewards or penalties, then adjusting to maximize total reward.

Data & AI

Unlike supervised learning, there are no labeled answers — the agent discovers good behavior through feedback over time. It must balance exploration (trying new things) with exploitation (using what already works).

It works much like training a dog with treats: good actions earn rewards, bad actions don't, and over many attempts the agent learns a strategy (called a policy). RL powers game-playing AIs and robotics, and helps fine-tune large language models.

Examples

Python — The agent repeatedly acts, observes a reward, and updates its policy.

# Sketch of the RL loop
state = env.reset()
for step in range(1000):
    action = agent.choose(state)
    state, reward, done = env.step(action)
    agent.learn(state, action, reward)
    if done:
        break

Learn it in a lesson

See also

Back to the Dictionary