What is the name of the problem of inferring the reward function an expert is optimizing purely from observations of their behavior, rather than being given the reward directly?

Show answer

Inverse reinforcement learning (IRL)

Inverse reinforcement learning flips the usual RL setup: instead of finding a policy that maximizes a known reward, it recovers the latent reward that best explains demonstrated behavior. It is useful when good behavior is easy to demonstrate but the underlying objective is hard to specify.

Also asked as:

What framework aims to recover the latent reward function that best explains an expert's demonstrated behavior?

Would you remember this in 6 months?

Trivia Forge uses spaced repetition so facts stick. Free, no signup needed.

Turn on Machine Learning practice →

More Machine Learning trivia

What is the name of the theorem stating that a feed-forward network with a single hidden layer and a nonpolynomial activation function can approximate any continuous function on a compact set to arbitrary accuracy?
What is the name of the value function that estimates the total reward an agent can expect to collect from a given state-action pair in Reinforcement Learning?
What is the standard name for the loss function usually optimized in logistic regression for binary classification?
What is the term for a language model's ability to perform a new task purely from instructions and examples provided in its prompt, without any updates to its weights?
What is the term for the learning setting in which the algorithm itself chooses which unlabeled examples to have labeled, typically the most uncertain or informative ones, to reach high accuracy with fewer labels?
What is the term for the simplest form of imitation learning, in which a policy is trained with supervised learning to directly map states to an expert's demonstrated actions?

See all 98 →