Imitation Learning Quiz

Answers and explanations appear as you take the quiz.

1. In imitation learning, how does an agent acquire its behavior?
   • Learning a policy by copying expert demonstrations
   • Maximizing a hand-designed reward through random exploration
   • Compressing a neural network to run on small devices
   • Generating images from text prompts
2. Behavioral cloning frames imitation learning as which kind of problem?
   • Unsupervised clustering
   • Supervised learning of state-to-action mappings
   • Adversarial generation
   • Reward shaping
3. What problem causes behavioral cloning to fail over long action sequences?
   • Overfitting to the reward function
   • Vanishing gradients in the encoder
   • Compounding error from distribution shift
   • Catastrophic forgetting of old classes
4. How does the DAgger algorithm address that weakness?
   • By removing the expert from the loop entirely
   • By increasing the discount factor
   • By adding random noise to all actions
   • By querying the expert on states the learner actually visits and aggregating that data
5. Why is imitation learning often preferred over reinforcement learning for tasks like driving?
   • Demonstrations are easier to provide than a correct reward function
   • It guarantees optimal behavior
   • It never needs any data
   • It runs faster at inference only
6. What is a realistic limitation of behavioral cloning when the agent reaches states outside the demonstrations?
   • It is guaranteed to improve beyond the expert
   • Its behavior is not guaranteed there because those states may be out of distribution
   • It can only produce random actions
   • It automatically queries an expert at every step
7. In one DAgger iteration, what does the learner do before requesting expert labels?
   • Deletes all earlier demonstrations
   • Trains only on randomly generated labels
   • Rolls out its current policy and records the states it actually visits
   • Replaces the expert policy permanently
8. After the expert labels the states visited by the current policy, what happens to that data?
   • It is used only as an evaluation metric
   • It replaces the original observation space
   • It is discarded after one gradient step
   • It is aggregated with previous labeled data for retraining