Program

Monday: Real-time Robot Learning

Morning:

9:00-9:30: Intro and Welcome: Rich Sutton: Real-time Reinforcement Learning
9:30-10:00: Thomas Degris: Continuous Actor-Critic
10:00-10:30:Todd Hester: A Real-Time Model-Based Reinforcement Learning Architecture for Robot Control

10:30-11:00 Break

11:00-11:30: Adam White: Multi-time-scale Nexting in a Reinforcement Learning Robot
11:30-12:00: Joseph Modayil: Horde: A Scalable Real-time Architecture for Learning Knowledge from Unsupervised Sensorimotor Interaction
12:00-12:30: Wouter Caarls: Parallel real-time reinforcement learning
12:30-1:00: Ashique Rupam Mahmood: On the Role of Step-Size Adaptation in Real-Time Reinforcement Learning

Evening:

7:00-8:00: Erik Schuitema: Real-time temporal difference learning on bipedal walking robot LEO
8:00-xx: Robot demos

Discussion: Learning on Robots (or other topics)

Tuesday:

Morning: Models

9:00-10:00: Alex Stoytchev: A Developmental Approach to Robotic Intelligence
10:00-10:30: George Konidaris: Autonomous Robot Skill Acquisition

10:30-11:00: break

11:00-11:30: Yuval Tassa: Fast Model Predictive Control generates real-time, reactive swimming
11:30-12:00: Tom Erez: Infinite-horizon model-predictive control for periodic domains with contact
12:00-12:30: John Roberts: Feedback Controller Parameterizations for Reinforcement Learning
12:30-1:00: Sylvie Ong: Efficient POMDP planning - Exploiting structure through belief space representations

Evening: Psychology

7:00-7:30: E. James Kehoe: Acquisition of Timing as Revealed in a Classically-Conditioned Protective Reflex
7:30-8:00: Elliot Ludvig: Real-time reinforcement learning in the brain
8:00-8:30: Francois Rivest: Learning timing in real-time using adaptive drift-diffusion processes

Discussion: Real-Time Learning in Nature

Wednesday

Morning: Scalable Learning

9:00-10:00: Doina Precup: Learning with Random Options
10:00-10:30
: Istvan Szita: Reinforcement learning with lame features

10:30-11:00
: Break

11:00-11:30
Hamid Maei: Toward lifting the curse-of-dimensionality in reinforcement learning
11:30-12:00
: Harm van Seijen: Exploiting Best-Match Equations for Efficient Reinforcement Learning
12:00-12:30
: Eduardo Alonso: Value Gradient Learning
12:30-1:00
: Yuri Grinberg: Limits of learning from a single trajectory

Evening: Break-out Discussions

Possible topics: Abstraction/Learning in high dimensional domains

Thursday

Morning: Knowledge & Representation

9:00-10:00: Rich Sutton: The problem of knowledge
10:00-10:30: David Silver: Question Answer Processes
10:30-11:00: Peng Zang: Making non-tabular representations more friendly

11:00-11:30: Break + Group Photo

11:30-12:00: Jeshua Bratman: Unsupervised Representation Learning: A Case Study in Computer Go
12:00-12:30: Phil Bachman: Informative Representations

12:30-1:00: Closing Discussion.
End of Barbados RL Workshop 2011
Ċ
Thomas Degris-Dard,
Mar 23, 2011, 8:30 AM
Ċ
Thomas Degris-Dard,
Mar 23, 2011, 8:32 AM
Ċ
Thomas Degris-Dard,
Jun 1, 2011, 8:46 PM
ć
Thomas Degris-Dard,
Mar 23, 2011, 11:13 AM
ć
Thomas Degris-Dard,
Mar 25, 2011, 10:44 AM
Ċ
Thomas Degris-Dard,
Mar 25, 2011, 10:45 AM
ć
Thomas Degris-Dard,
Mar 23, 2011, 8:57 AM
Ċ
Thomas Degris-Dard,
Mar 23, 2011, 8:35 AM
Ċ
MaeiTalk.pdf
(1562k)
Thomas Degris-Dard,
Mar 23, 2011, 11:15 AM
Ċ
Thomas Degris-Dard,
Mar 25, 2011, 10:46 AM
Ċ
Thomas Degris-Dard,
Mar 23, 2011, 8:38 AM
Ċ
Thomas Degris-Dard,
Mar 23, 2011, 12:15 PM
ċ
SchuitemaTalk.ppt
(16041k)
Thomas Degris-Dard,
Apr 1, 2011, 11:46 PM
ċ
SeijenTalk.ppt
(1065k)
Thomas Degris-Dard,
Mar 27, 2011, 5:27 PM
Ċ
Thomas Degris-Dard,
Mar 25, 2011, 10:52 AM
Ċ
Thomas Degris-Dard,
Mar 21, 2011, 4:07 PM
Ċ
Thomas Degris-Dard,
Apr 1, 2011, 11:47 PM
Ċ
Thomas Degris-Dard,
Mar 25, 2011, 10:47 AM
Ċ
Thomas Degris-Dard,
Mar 25, 2011, 10:51 AM
Comments