Sciweavers

Free Online Productivity Tools i2Speak i2Symbol i2OCR iTex2Img iWeb2Print iWeb2Shot i2Type iPdf2Split iPdf2Merge i2Bopomofo i2Arabic i2Style i2Image i2PDF iLatex2Rtf Sci2ools

12

AGI
2015

favoriteEmaildiscussreport

13views Artificial Intelligence» more AGI 2015»

Using Localization and Factorization to Reduce the Complexity of Reinforcement Learning

8 years 13 days ago

Using Localization and Factorization to Reduce the Complexity of Reinforcement Learning

Download agi-conf.org

General reinforcement learning is a powerful framework for artiﬁcial intelligence that has seen much theoretical progress since introduced ﬁfteen years ago. We have previously provided guarantees for cases with ﬁnitely many possible environments. Though the results are the best possible in general, a linear dependence on the size of the hypothesis class renders them impractical. However, we dramatically improved on these by introducing the concept of environments generated by combining laws. The bounds are then linear in the number of laws needed to generate the environment class. This number is identiﬁed as a natural complexity measure for classes of environments. The individual law might only predict some feature (factorization) and only in some contexts (localization). We here extend previous deterministic results to the important stochastic setting.

Peter Sunehag, Marcus Hutter

Real-time Traffic

AGI 2015 | Artificial Intelligence |

claim paper

Related Content

» On the asymptotic equivalence between differential Hebbian and temporal difference learnin...

» Automatic Feature Selection for ModelBased Reinforcement Learning in Factored MDPs

» Adaptive Routing for Sensor Networks using Reinforcement Learning

» Learning to play Tetris applying reinforcement learning methods

» Combining manual feedback with subsequent MDP reward signals for reinforcement learning

» Reinforcement of Local Pattern Cases for Playing Tetris

» Localizing Search in Reinforcement Learning

» Active Learning for Reward Estimation in Inverse Reinforcement Learning

» Reinforcement learning by rewardweighted regression for operational space control

» RLTOPS An Architecture for Modularity and ReUse in Reinforcement Learning

Post Info
More Details (n/a)

Added	13 Apr 2016
Updated	13 Apr 2016
Type	Journal
Year	2015
Where	AGI
Authors	Peter Sunehag, Marcus Hutter

Comments (0)