Imitation learning uses examples from a human, another agent or a recorded policy to teach behavior. The system learns a mapping from observations to actions or infers the objective behind the demonstrations.
One-shot task imitation aims to reproduce a new behavior from a single demonstration. That is challenging because the agent must separate the essential task from incidental details and adapt it to a new situation. Errors in the demonstration can also be copied, so evaluation and corrective feedback remain important.
Acronyms and aliases
IL acronymlearning from demonstration synonym
Related terms
Frequently asked questions
How is imitation learning different from reinforcement learning?
Imitation learning starts from demonstrated behavior. Reinforcement learning improves actions using reward feedback, though practical systems often combine both methods.
What is one-shot imitation learning?
It is the attempt to learn or reproduce a task from one demonstration. The system must generalize the demonstrated intent rather than merely replaying the exact motion.
