Fixed vs. dynamic sub-transfer in reinforcement learning (2002) [4 citations — 1 self]
by James L. Carroll
In Proceedings of the International Conference on Machine Learning and Applications
http://aml.cs.byu.edu/papers/dynamic.ps
Add To MetaCart
Abstract:
We survey various transfer methods in Q-learning, a type of reinforcement learning, and present a variation on fixed sub-transfer which we call dynamic sub-transfer. We describe the pros and cons of dynamic sub-transfer as compared with the other transfer methods, and we describe qualitatively the situations where this method would be preferred over the fixed version of sub-transfer. 1
Citations
| 137 | Multitask learning – Caruana - 1997 |
| 80 | Finding structure in reinforcement learning – Thrun, Schwartz - 1995 |
| 53 | The Behavior of Organism: An Experimental Analysis – Skinner - 1939 |
| 31 | Lifelong robot learning – Thrun, Mitchell - 1993 |
| 16 | Incorporating prior knowledge and previously learned information into reinforcement learning agents – Dixon, Malak, et al. - 2000 |
| 9 | Bounding the suboptimality of reusing subproblems – Bowling, Veloso - 1999 |
| 9 | Reusing learned policies between similar problems – Bowling, Veloso - 1998 |
| 5 | Automated shaping as applied to robot navigation – Peterson, Owens, et al. - 2001 |
| 4 | Memory-guided exploration in reinforcement learning – Carroll, Peterson, et al. - 2001 |
| 1 | Multitask learning," in Learning to Learn, Lorien Pratt – Caruana - 1998 |
