Learner-Aware Teaching: Inverse Reinforcement Learning with Preferences and Constraints
Abstract
Inverse reinforcement learning (IRL) enables an agent to learn complex behavior by observing demonstrations from a (near-)optimal policy. The typical assumption is that the learner's goal is to match the teacher’s demonstrated behavior. In this paper, we consider the setting where the learner has its own preferences that it additionally takes into consideration. These preferences can for example capture behavioral biases, mismatched worldviews, or physical constraints. We study two teaching approaches: learner-agnostic teaching, where the teacher provides demonstrations from an optimal policy ignoring the learner's preferences, and learner-aware teaching, where the teacher accounts for the learner’s preferences. We design learner-aware teaching algorithms and show that significant performance improvements can be achieved over learner-agnostic teaching.
Cite
Text
Tschiatschek et al. "Learner-Aware Teaching: Inverse Reinforcement Learning with Preferences and Constraints." Neural Information Processing Systems, 2019.Markdown
[Tschiatschek et al. "Learner-Aware Teaching: Inverse Reinforcement Learning with Preferences and Constraints." Neural Information Processing Systems, 2019.](https://mlanthology.org/neurips/2019/tschiatschek2019neurips-learneraware/)BibTeX
@inproceedings{tschiatschek2019neurips-learneraware,
title = {{Learner-Aware Teaching: Inverse Reinforcement Learning with Preferences and Constraints}},
author = {Tschiatschek, Sebastian and Ghosh, Ahana and Haug, Luis and Devidze, Rati and Singla, Adish},
booktitle = {Neural Information Processing Systems},
year = {2019},
pages = {4145-4155},
url = {https://mlanthology.org/neurips/2019/tschiatschek2019neurips-learneraware/}
}