A Survey on Methods and Applications of Deep Reinforcement Learning

PhD Qualifying Examination


Title: "A Survey on Methods and Applications of Deep Reinforcement 
Learning"

by

Mr. Siyi LI


Abstract:

While perception tasks such as visual image classification and object 
detection play an important role in human intelligence, the more 
sophisticated tasks built upon them that involve decision and planning 
require an even higher level of intelligence. The past few years have seen 
major advances in many low-level perceptual supervised learning problems 
by using deep learning models. For higher-level tasks, however, 
reinforcement learning offers a more powerful and flexible framework for 
the general sequential decision making problem. While reinforcement 
learning has achieved some successes in a variety of domains, their 
applicability has previously been limited to domains with low-dimensional 
state spaces. To derive efficient and powerful feature representations of 
the environment, it is naturally desirable to incorporate deep learning to 
the reinforcement learning domains, which we call deep reinforcement 
learning. In this survey, we start from research on general reinforcement 
learning methods. We then review the recent advances in deep reinforcement 
learning, including both the methods and its applications on game playing 
and robotics control. Finally, we discuss some possible research issues.


Date:			Friday, 13 January 2017

Time:                  	3:00pm - 5:00pm

Venue:                  Room 3494
                         Lifts 25/26

Committee Members:	Prof. Dit-Yan Yeung (Supervisor)
 			Dr. Raymond Wong (Chairperson)
 			Prof. James Kwok
 			Prof. Nevin Zhang


**** ALL are Welcome ****