AIF-C01 Practice Question 21: Reinforcement Learning Reward
AWS Certified AI Practitioner (AIF-C01) · Test 3 · question 1 of 10
A robotics team trains a warehouse picking robot inside a simulator. Nobody labels the "correct" move for each situation. Instead, the robot tries an action, the simulator returns a score (positive when an item is placed in the right bin, negative when it drops one), and the robot gradually adjusts its behaviour to earn more points over a full run. Which machine learning paradigm does this describe?
Show answer and explanation
Answer:
Reinforcement learning (RL) is the paradigm in which an agent interacts with an environment, takes actions, observes the resulting state, and receives a numeric reward. The agent is never told which action was correct; it learns a policy that maximizes cumulative reward over the long term. The warehouse robot, the simulator, the moves and the points map exactly onto agent, environment, actions and reward signal (option C).
- A is wrong because supervised learning needs an external supervisor supplying labeled examples, each input paired with its correct answer. A score for an action taken is a reward, not a label saying what the right action was. - B is wrong because unsupervised learning finds structure inside unlabeled data, such as clusters or embeddings. Here there is a reward signal driving the learning, which is the one thing unsupervised learning does not have. - D is wrong because semi-supervised learning mixes a small labeled dataset with a large unlabeled one. The scenario has no labels at all, only feedback on actions already taken.
Reference: Use Reinforcement Learning with Amazon SageMaker AI