ICML Poster Environment Design for Inverse Reinforcement Learning

Poster

Environment Design for Inverse Reinforcement Learning

Thomas Kleine Buening · Victor Villin · Christos Dimitrakakis

Hall C 4-9 #1215

[ Abstract ] [ Paper PDF ]

[ Poster]

Thu 25 Jul 4:30 a.m. PDT — 6 a.m. PDT

Oral presentation: Oral 6F Experimental Design and Simulation
Thu 25 Jul 7:30 a.m. PDT — 8:30 a.m. PDT

Abstract:

Learning a reward function from demonstrations suffers from low sample-efficiency. Even with abundant data, current inverse reinforcement learning methods that focus on learning from a single environment can fail to handle slight changes in the environment dynamics. We tackle these challenges through adaptive environment design. In our framework, the learner repeatedly interacts with the expert, with the former selecting environments to identify the reward function as quickly as possible from the expert’s demonstrations in said environments. This results in improvements in both sample-efficiency and robustness, as we show experimentally, for both exact and approximate inference.

Chat is not available.