Deep Reinforcement Learning

  • Aggarwal C
Citations of this article
Mendeley users who have this article in their library.
Get full text


Reinforcement Learning (RL) is a type of Machine Learning algorithms and it enables the agent to determine the ideal behavior from its own experience. From robot control to autonomous navigation, Reinforcement Learning algorithms have been applied to address increasing difficult problems. In recent studies, a number of papers have shown great success of RL in the field of production control, finance, scheduling, communication and auto vehicle control. However, in most cases, the performance of these algorithms heavily rely on the quality of the handcrafted features. This drawback limits the application scope of traditional Reinforcement Learning algorithms, since some problems have high dimensional state space and are difficult to hand-engineered. For instance, it is a long standing challenge for traditional RL algorithms to process high dimensional sensory data like vision and voice. In 2006, the Deep Learning (DL) algorithms were established and have been further developed in recent years. The Convolutional Neural Network is one of the Deep Learning models that could extract high dimensional features direct from the raw pixels, and have been successfully applied in computer vision. It is nature for us to think whether the traditional Reinforcement Learning algorithms could benefit from it.




Aggarwal, C. C. (2018). Deep Reinforcement Learning. In Neural Networks and Deep Learning (pp. 373–417). Springer International Publishing.

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Save time finding and organizing research with Mendeley

Sign up for free