Skip to content
/ MGRL Public

Multigroup Reinforcement learning for Electrolyte Repletion

Notifications You must be signed in to change notification settings

bee-hive/MGRL

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MGRL

Multigroup Reinforcement learning for Electrolyte Repletion

Overview

This repository presents a novel approach in healthcare data analytics, focusing on off-policy reinforcement learning methods for treatment policies. Our work is particularly tailored to the complex and diverse nature of patient populations, where individuals often present different chronic conditions requiring personalized treatment strategies.

Introduction

Traditional reinforcement learning models in healthcare often overlook the heterogeneity within patient populations. To address this, our research implements a multi-group Gaussian process regression model within a fitted Q-iteration framework. This approach enables us to:

  • Model diverse patient subgroups accurately.
  • Tailor optimal treatment policies to each subgroup.
  • Estimate these functions across the entire patient population.

Application

We apply our Multi-Group Reinforcement Learning (MGRL) framework to the critical problem of formulating optimal treatment policies for electrolyte repletion in patients with pre-existing medical conditions.

Key Features

  • Multi-Group Gaussian Process (MGGP) Regression Models: Allow for precise modeling of different patient subgroups.
  • Fitted Q-Iteration Framework: Ensures robust policy development and adaptation for each subgroup.
  • Whole Population Estimation: Provides a comprehensive view of the patient population, enhancing the treatment policy's overall effectiveness.

About

Multigroup Reinforcement learning for Electrolyte Repletion

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published