Policy Improvement for POMDPs Using Normalized Importance Sampling

dc.creator	Shelton, Christian R.
dc.date	2004-10-20T20:50:06Z
dc.date	2004-10-20T20:50:06Z
dc.date	2001-03-20
dc.date.accessioned	2013-10-09T02:48:35Z
dc.date.available	2013-10-09T02:48:35Z
dc.date.issued	2013-10-09
dc.identifier	AIM-2001-002
dc.identifier	CBCL-194
dc.identifier	http://hdl.handle.net/1721.1/7218
dc.identifier.uri	http://koha.mediu.edu.my:8181/xmlui/handle/1721
dc.description	We present a new method for estimating the expected return of a POMDP from experience. The estimator does not assume any knowle ge of the POMDP and allows the experience to be gathered with an arbitrary set of policies. The return is estimated for any new policy of the POMDP. We motivate the estimator from function-approximation and importance sampling points-of-view and derive its theoretical properties. Although the estimator is biased, it has low variance and the bias is often irrelevant when the estimator is used for pair-wise comparisons.We conclude by extending the estimator to policies with memory and compare its performance in a greedy search algorithm to the REINFORCE algorithm showing an order of magnitude reduction in the number of trials required.
dc.format	4576001 bytes
dc.format	768071 bytes
dc.format	application/postscript
dc.format	application/pdf
dc.language	en_US
dc.relation	AIM-2001-002
dc.relation	CBCL-194
dc.title	Policy Improvement for POMDPs Using Normalized Importance Sampling

Files in this item

Files	Size	Format	View
There are no files associated with this item.