Adaptive Clinical Treatments and Reinforcement Learning for Automatic Disease diagnosis

Abstract

Machine learning models are taught how to make a series of decisions depending on a set of inputs in reinforcement learning. The agent learns how to accomplish a goal in an unexpected, maybe complex environment. Reinforcement learning places artificial intelligence in a game-like environment. It solves the problem by trial and error. Artificial intelligence is rewarded or punished based on its actions. Its purpose is to maximize the amount of money paid out in total. In addition to providing the game's rules, the designer does not give any feedback or recommendations on how to win the model. To maximize reward, the model must determine the optimum way to do a job, beginning with purely random trials and progressing to complex techniques and superhuman abilities. Reinforcement learning, with its power of search and diversity of trials, is likely the most effective strategy for hinting at a system's originality. Unlike humans, AI can learn from thousands of concurrent gameplays if a reinforcement learning algorithm is run on sufficiently efficient computer infrastructure.