Repeated Inverse Reinforcement Learning

Abstract

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.

arXiv commentThe first two authors contributed equally to this work. The paper appears in NIPS 2017

Cited by in corpus (15)

• Deep Reinforcement Learning: An Overview
• Scalable agent alignment via reward modeling: a research direction
• Learner-aware Teaching: Inverse Reinforcement Learning with Preferences and Constraints
• Open Problems in Cooperative AI
• Learning Constraints from Demonstrations
• Multi-task Maximum Entropy Inverse Reinforcement Learning
• A Framework and Method for Online Inverse Reinforcement Learning
• Efficient Model-Free Reinforcement Learning Using Gaussian Process
• Identifiability in inverse reinforcement learning
• Active Inverse Reward Design
• Curriculum Design for Teaching via Demonstrations: Theory and Applications
• Learning Parametric Constraints in High Dimensions from Demonstrations
• Interaction-limited Inverse Reinforcement Learning
• Learning from an Exploring Demonstrator: Optimal Reward Estimation for Bandits
• Value Alignment Verification