Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1142364 | Operations Research Letters | 2015 | 5 Pages |
Abstract
In a ranking and selection problem with independent Bernoulli populations we study the knowledge-gradient policy. We show that this policy is inefficient and propose three alternative rules that have superior performance. In particular we show that one of the considered policies–called KG(∗∗) in the literature–is numerically difficult and expensive to evaluate and it is possible to derive a much simpler policy that gives comparable results.
Related Topics
Physical Sciences and Engineering
Mathematics
Discrete Mathematics and Combinatorics
Authors
Bogumił Kamiński,