Repeated Inverse Reinforcement Learning

By Kareem Amin, Nan Jiang, Satinder Singh. 2017

78 citations recorded by the source corpus.

We introduce a novel repeated Inverse Reinforcement Learning problem: the agent has to act on behalf of a human in a sequence of tasks and wishes to minimize the number of tasks that it surprises the human by acting suboptimally with respect to how the human would have acted. Each time the human is surprised, the agent is provided a demonstration of the desired behavior by the human. We formalize this problem, including how the sequence of tasks is chosen, in a few different ways and provide some foundational results.

Subfields: Artificial Intelligence, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Sergey Levine
2. Tom Everitt
3. Jan Leike
4. Shane Legg
5. Pieter Abbeel
6. Justin Fu
7. David Krueger
8. Chelsea Finn

In Reinforcement Learning

1. Sergey Levine
2. Pieter Abbeel
3. Justin Fu
4. Jan Leike
5. Tom Everitt
6. Chelsea Finn
7. Shane Legg
8. David Krueger

Citations in the corpus

• Scalable agent alignment via reward modeling: a research direction (2018, background): “If we assume the human is fully rational and the agent can design a sequence of tasks for the human, then the reward function can be identified (Amin et al., 2017).”
• Learning Robust Rewards with Adversarial Inverse Reinforcement Learning (2017, methodology): “Amin et al. (2017) consider learning reward functions which generalize to new tasks given multiple training tasks.”