Title Of the Talk: Reinforcement Learning Beyond Rewards
Speaker: Prof B Ravindran
Host Faculty: Dr. Vineeth N Balasubramanian
Date &Time: Friday, 29th Nov 2019 2:30 PM
Deep Reinforcement Learning methods have achieved significant successes recently by marrying the representation learning power of deep networks and the control learning abilities of RL. This success has opened up new lines of research and revived old ones in the RL community. I will talk about two pieces of work that go beyond reward-based RL. The first is Macro and Micro Curriculum (MaMiC) [AAMAS 2019] that discusses curriculum strategies at two levels; one relating to the sequence of sub-tasks that are required to solve an overall task using information from an expert, while the other relating to solving each given sub-task through achieving artificial sub-goals that gradually increase in difficulty. The second is Successor Options [IJCAI 2019] - on discovering hierarchical structure in problems by exploiting the properties of successor representations.
Prof. Ravindran is the head of the Robert Bosch Centre for Data Science and Artificial Intelligence (RBC-DSAI) at IIT Madras and a professor and the Mindtree Faculty Fellow in the Department of Computer Science and Engineering. He is also the co-director of the reconfigurable and intelligent systems engineering (RISE) group at IIT Madras. He received his PhD from the University of Massachusetts, Amherst and his Master’s in research degree from Indian Institute of Science, Bangalore. He has nearly two decades of research experience in machine learning and specifically reinforcement learning. He has held visiting positions at the Indian Institute of Science, Bangalore, India and University of Technology, Sydney, Australia. He is also one of the founding executive committee members of the India chapter of ACM SIGKDD (IKDD) and is currently serving as the president of the chapter. His research interests are centred on learning from and through interactions and span the areas of complex network analysis and reinforcement learning.
Friday, 29th Nov 2019 2:30 PM