The use of reinforcement learning methods in medical problems

Abstract

In this article the features of the modern reinforcement learning methods development for the medical tasks are discussed. Reinforcement learning methods are a popular machine learning tool used in the problems of finding optimal patient treatment strategies, personalized medicine, as well as interactive patient monitoring systems. One of the important task is to choose the optimal reinforcement learning algorithm from a variety of currently existing methods that have their own application specifics, advantages and disadvantages. This article is devoted to the analysis of the algorithmic apparatus of the most popular reinforcement learning methods and contains examples of models and results of the methods under consideration in the context of the problem of finding optimal treatment regimens for cardiac patients.