How is reinforcement learning (RL) different from supervised learning?

• A. RL learns by trial-and-error through feedback from actions, while supervised learning learns from labeled examples.
• B. RL uses labeled data only; supervised uses reward signals.
• C. RL cannot handle sequential tasks; supervised cannot.
• D. RL is a form of unsupervised learning.

Answer: (A)

Reinforcement learning centers on an agent that learns by interacting with an environment, taking actions, and receiving rewards that reflect how good those actions were over time. The key is trial-and-error and credit assignment: the agent must figure out which actions early in a sequence led to long-term rewards and adjust its behavior accordingly. Supervised learning, on the other hand, uses a fixed set of labeled examples that pair inputs with correct outputs. The model learns a direct mapping by minimizing error on those examples, without needing to consider how actions unfold over time or affect future states.

This distinction matters: RL handles sequential decision making with potentially delayed rewards, where the best action depends on future consequences. Supervised learning assumes static data and aims to predict labels for new inputs based on patterns learned from labeled pairs. There are advanced hybrids, like learning from demonstrations, but the core difference remains reward-based learning from interactions versus label-based learning from fixed examples.