Potentials of reinforcement learning in contemporary scenarios

Abstract

This paper reviews the present applications of reinforcement learning in five major spheres including mobile autonomy, industrial autonomy, finance and trading, and gaming. The application of reinforcement learning in real time cannot be overstated, it encompasses areas far beyond the scope of this paper, including but not limited to medicine, health care, natural language processing, robotics and e-commerce. Contemporary reinforcement learning research teams have made remarkable progress in games and comparatively less in the medical field. Most recent implementations of reinforcement learning are focused on model-free learning algorithms as they are relatively easier to implement. This paper seeks to present model-based reinforcement learning notions, and articulate how model-based learning can be efficient in contemporary scenarios. Model based reinforcement learning is a fundamental approach to sequential decision making, it refers to learning optimal behavior indirectly by learning a model of the environment, from taking actions and observing the outcomes that include the subsequent sate and the instant reward. Many other spheres of reinforcement learning have a connection to model-based reinforcement learning. The findings of this paper could have both academic and industrial ramifications, enabling individual.