Program



Slides from the presentations

Sunday, April 28 

Morning: 

Where do models come from?


9:00 - 9:45           Rich Sutton - Planning in RL

9:45 -10:30          Byron Boots - Hilbert Space Embeddings of Predictive State Representations


10:30 -11:00      Break


11:00 -11:45        Erik Talvitie - Learning to make predictions in high-dimensional, partially observable domains

11:45 -12:15        Scott Sanner -Symbolic Dynamic Programming for Hybrid MDPs

12:15 -12:45        Marc Bellemare - Pixels and Priors: Learning a Generative Model of Atari 2600 Games

Evening:

7:00 - 8:00           Mausam and Andrey Kolobov -- An Overview of MDP Planning Research at ICAPS [Lecture Room]

8:10 - 9:30           Discussion on Representations for Planning led by Scott Sanner & Michael Littman [Dinnner Area]

Monday, April 29

Morning: 

Temporal abstraction


9:00 - 9:30     Doina Precup - Planning to explore using options

9:30 - 10:00     Dave Silver - Compositional Planning Using Optimal Option Models

10:00 -10:30     Tom Schaul - Better generalization with forecasts


10:30 -11             Break


Guiding search


11:00 -11:30     Tobias Jung - Optimized Lookahead Tree Policies: A Bridge Between Lookahead Tree Policies and Direct Policy Search

11:30 -12:00     Harm van Seijen - Efficient Prioritized Sweeping with Exact Bellman Error Ordering

12:00 -12:30     Csaba Szepesvari - Optimistic optimization algorithms for planning in MDPs I

12:30 - 1:00         Andras Gyorgy - Optimistic optimization algorithms for planning in MDPs II

Evening: 

7:00 - 7:30     Elliot Ludvig - Making decisions by replaying the past

7:30 - 8:15     John O’Doherty - Neural correlates of arbitration between model-based and model-free reinforcement-learning


8:15 - 8:30           Break


8:30 - 9:00           Matt Botvinick - Optimal behavioral hierarchy

9:00 - 9:30     Yuan Chang - Learning what’s relevant in a largely irrelevant world: The role of selective attention in representation learnings

Tuesday, April 30

Morning: 

POMDPs & Traditonal Planning


9:00 - 9:30     Richard Dearden - Replanning in POMDPs in Response to Unexpected Events

9:30 - 10:00     Pascal Poupart - Constrained POMDPs

10:00 - 10:30     Dan Weld - Specifying Adaptive Programs – a Challenge for RL


10:30 - 11:00     Break


11:00 - 11:30       Subramanian Ramamoorthy - Clustering Markov Decision Processes for a Lifelong Learning Agent

11:30 - 12:00     Mausam - Automatic Basis Function Construction for Goal-directed MDPs

12:00 - 12:30     Andrey Kolobov - Pitfalls of Goal-Oriented Planning Under Uncertainty

Evening:

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

Wednesday, May 1

Morning: 

Applications


9:00 - 9:30     Joseph Modayil - Planning with learned models on robots

9:30 - 10:00     Patrick Pilarski - Real-time Control with Temporally Extended Predictions: A Sensorimotor Approach to Planning?

10:00 - 10:30     Wouter Caarls - Parallel DYNA


10:30 - 11:00     Break



11:00 -11:30     Michael Bowling - Arcade Learning Environment

11:30 -12:00     Samuel Barrett - Planning to Cooperate with Unknown Teammates

12:00 -12:30     Russ Greiner - Learning to help patients manage diabetes

Evening: 

TBD