Non-Markovian Reward Modelling from Trajectory Labels via Interpretable Multiple Instance LearningDownload PDF

Published: 31 Oct 2022, Last Modified: 12 Jan 2023NeurIPS 2022 AcceptReaders: Everyone
Keywords: reward modelling, reinforcement learning, multiple instance learning, interpretability
TL;DR: We generalise reward modelling for reinforcement learning to handle non-Markovian rewards, and propose new interpretable multiple instance learning models for this problem.
Abstract: We generalise the problem of reward modelling (RM) for reinforcement learning (RL) to handle non-Markovian rewards. Existing work assumes that human evaluators observe each step in a trajectory independently when providing feedback on agent behaviour. In this work, we remove this assumption, extending RM to capture temporal dependencies in human assessment of trajectories. We show how RM can be approached as a multiple instance learning (MIL) problem, where trajectories are treated as bags with return labels, and steps within the trajectories are instances with unseen reward labels. We go on to develop new MIL models that are able to capture the time dependencies in labelled trajectories. We demonstrate on a range of RL tasks that our novel MIL models can reconstruct reward functions to a high level of accuracy, and can be used to train high-performing agent policies.
Supplementary Material: pdf
18 Replies