Tuesday: Views of Model Based Reinforcement Learning (09h00):Chair: Richard Sutton09h00-09h40: Using hierarchical Bayes to learn models and plan in them David Wingate (30min talk) 09h40-10h20: Lifted reasoning (for RL and beyond) Geoff Gordon (30min talk) plus Geoff recommends this tutorial on lifted probabilistic representations by David Poole. 10h20-10h40: Break 10h40-11h00: Temporal Abstractions for Knowledge Representation (Anna Koop) (20min talk) 11h00-11h20: Modeling the world with multi-layer predictrons (David Silver) (20min talk) 11h20-11h40: Break 12h20-13h00: Planning with an untrustworthy model Michael Bowling (30min talk)
Evening (19h00):Chair: Doina Precup9h40-20h20: Do animals learn models of their environment Elliot Ludvig (30min talk) 20h20-20h40: Break 20h40-21h20: Cognitive neuroscience and goal-directed (read model-based decision making Matthew Botvinick (30min talk) 19h00-19h30: Where do rewards come from? Richard Lewis (30min talk) 21h20-22h00: Open Discussion Wednesday:
Planning (9h00)
:Chair: Michael Bowling09h00-09h40: Planning and RL: A tale of two worlds Prasad Tadepalli (30min talk) 09h40-10h00: Learning and planning with partial models (Neville Mehta) (20min talk) 10h00-10h20: Optimistic planning of deterministic systems (Jean-François Hren) (20min talk) 10h20-10h40: Discussion 10h40-11h00: Break Chair: Martijn van OtterloPlanning:11h00-11h40: Random thoughts Csaba Szepesvári (30min talk) 11h40-12h00: Fitted Natural Actor-Critic: A new algorithm for continuous state-action MDPs (Francisco Melo) (20min talk) Representations: 12h00-12h20: Embedding 101: Reconstructing nonlinear dynamical systems from time series data (Jordan Frank) 12h20-12h40: Manifold Embeddings for Reinforcement Learning with Partial Observability (Keith Bush) 12h40-13h00: Discussion Evening (19h00-22h00):Open discussion
Thursday:
Temporal abstraction (09h00):Chair: Prasad Tadepalli09h00-09h40: Doina Precup (30min talk) 09h40-10h10: Linear Option Models (Jonathan Sorg) (20min talk) 10h10-10h40: Optimal Policy Switching Algorithms (Gheorghe Comanici) (20min talk) 10h40-10h50: Break
Model learning:Chair: Csaba Szepesvári10h50-11h20: Tools for learning RL representations (Carlos Diuk) (20min talk) 11h20-11h50: Gradient-descent methods for temporal difference learning with linear function approximation (Hamid Maei) (20min talk) 11h50-12h20: Transfer Learning in Related Reinforcement Learning Tasks (Alessandro Lazaric) (20min talk) 12h20-12h50: Model learning in strongly structured environments (Gabor Bartok) (20min talk)
Evening: Strategies for progress (19h00): Chair: Dale Schuurmans19h00-19h20: Critterbot (Thomas Degris-Dard) (20min talk) 19h20-19h40: The RLAI Robotic Simulator (Marc Bellemare) (20min talk) 19h40-20h00: Discussion 20h00-20h30: RL Community Tools State of the Union AND A Trail Towards More General Empirical Reinforcement Learning (Brian Tanner) 20h30-22h00: Discussion
Friday:
Representations (9h00): Chair: David Wingate09h00-09h20: Generalized Model Learning in Reinforcement Learning (Todd Hester) (20min talk) 09h20-09h40: Model-based Bayesian Reinforcement Learning with Adaptive State Aggregation (Cosmin Paduraru) (20min talk) 09h40-10h00: Discussion 10h00-10h30: Break Chair: Geoff Gordon10h30-11h00: Learning Approximate Representations of Partially Observable Systems (Monica Dinculescu) (20min talk) 11h00-11h30: OOMs, PSRs, S-MAs and a statistically efficient learning algorithm (Michael Thon) (20min talk) 11h30-12h00: Maintaining the values of predictive features without a model (Erik Talvitie)
(20min talk) 12h00-12h30: Error-Bars for Bayesian Net Inference (Yasin Abbasi) (20min talk) |