Identifying Decision Points for Safe and Interpretable Reinforcement Learning in Hypotension Treatment
Machine Learning
2021-01-12 v1
Abstract
Many batch RL health applications first discretize time into fixed intervals. However, this discretization both loses resolution and forces a policy computation at each (potentially fine) interval. In this work, we develop a novel framework to compress continuous trajectories into a few, interpretable decision points --places where the batch data support multiple alternatives. We apply our approach to create recommendations from a cohort of hypotensive patients dataset. Our reduced state space results in faster planning and allows easy inspection by a clinical expert.
Cite
@article{arxiv.2101.03309,
title = {Identifying Decision Points for Safe and Interpretable Reinforcement Learning in Hypotension Treatment},
author = {Kristine Zhang and Yuanheng Wang and Jianzhun Du and Brian Chu and Leo Anthony Celi and Ryan Kindle and Finale Doshi-Velez},
journal= {arXiv preprint arXiv:2101.03309},
year = {2021}
}
Comments
NeurIPS 2020 Machine Learning for Health (ML4H) Workshop