Program

Monday, April 27 

Morning: 

9:00 - 9:45           Doina Precup - Beyond individual algorithms: Computational architectures for reinforcement learning in AI


9:45 - 10:00         Break


10:00 - 10:45        Jan Peters: Putting it all together: An Integrated Architecture for Motor Skill Learning

10:45 - 11:15       Tom Schaul: Universal Value Function Approximators


11:15 - 11:30      Break


11:30 - 12:00        Anna Harutyunyan: Parallel reward shaping architectures

12:00 - 12:30        Joseph Modayil: Integrated architectures for function approximation

12:30 - 1:00          Matteo Leonetti: Robust and Adaptive Planning through Reinforcement Learning

Evening:

7:00 - 8:00           George Konidaris (via Skype): The Symbol Strikes Back: Building True Abstraction Hierarchies from the Bottom Up

8:00 - 9:00           Discussion

Tuesday, April 28

Morning: 

9:00 - 9:45            Elliot Ludvig: Learning and Choosing through Replaying Memories

9:45 - 10:15          Kevin Miller:

Evidence of Multiple Systems Guiding Rodent Behavior in a Two-Armed Bandit Task


10:15 - 10:30        Break


10:30 - 11:15        Giovanni Pezzulo: Planning in the Rat Hippocampus

11:15 - 11:45        Harm van Seijen: A Deeper Look at Planning as Learning from Replay


11:45 - 12:00        Break


12:00 - 12:30        Daniel Rasmussen: Biological neural modelling of hierarchical reinforcement learning

12:30 - 1:00         James McGlashan: Learning About Reward Functions

Evening: 

7:00 - 8:00 PM     Hado van Hasselt and Arthur Guez: Combining Reinforcement Learning with Deep Learning


8:00 - 9:30 PM     Discussions.


Topic 1: Recombination or Reinvention (aka Are we done yet?)

Topic 2: Challenge problems from the field

Wednesday, April 29

Morning: 

9:00 - 9:45         Borja Balle: PSR learning and Planning 9:45 - 10:15       Clement Gehring: Approximate Linear Successor Representations 10:15 - 10:45      Martha White: Local partitioned kernels for continuous state and action spaces


10:45 - 11:15     Break

11:15 - 11:45     Adam White: Experiences trying to put some of it together 11:45 - 12:15      Peter Sunehag: Slate-action Reinforcement Learning 12:15 - 12:45      Pierre-Luc Bacon: Learning recognizers

Evening:

7:00-9:30             Break-out Group Discussions

Thursday, April 30

Morning: 

9:00 - 9:45        Patrick Pilarski: Solving Artificial Intelligence (AI) with help from Intelligence Amplification (IA) 9:45 - 10:15      Ann Edwards: Toward the Shared Control of a Robotic Arm 10:15 - 10:45     Craig Sherstan: A Collaborative Approach to the Simultaneous Multi-joint Control of a Prosthetic Arm


10:45 - 11:00     Break

11:00 - 12:00     Rich Sutton: Emphatic Temporal-Difference Learning


End Barbados Workshop 2015