Talk:Gittins index

Untitled
I don't think the discussion regarding the selection of technologies is relevant to the Gittins index itself. Investing in technologies is not really parallel to the N-armed bandit problem except at a superficial and probably artificial level. 165.125.144.16 (talk) 19:46, 11 September 2009 (UTC)


 * I think you are right, but this example appears often in relation with the Multi-armed bandit process. Indeed what one does in the selection of technologies is to gather information in a bayesian way and the states of the process are probability distributions, which are called belief state.  There is an algorithm for computing the Gittins index in such a setup but it is quite complex and I agree that it would be just misleading in an introductory example. Pychron (talk) 00:42, 28 November 2009 (UTC)

Article's name
Shouldn't Gittins' have an apostrophe? Henriqueroscoe (talk) 14:48, 11 June 2010 (UTC)
 * Well, in the literature it does not, example Pychron (talk) 19:44, 18 June 2010 (UTC)
 * Editor Pychron is correct that the literature refers to the "Gittins index", just as mathematical literature referes to Hilbert space, Hardy space, Haar system, Haar measure (Hardy-Haar-Haar), Wald test, etc. Mathematical and scholarly practice differ from the recommendation of Strunk and White, who prescribe an apostrophe for forming the possessive for proper names (save Jesus and Moses). Cheers, Kiefer.Wolfowitz (talk) 10:20, 19 June 2010 (UTC)

Correct interpretation of Exploration vs Exploitation
The section "terminology" mentions a Bernoulli process with unknown probabilities. I would remove that reference: the basic Gittins index deals with processes whose probabilistic structure is known. The classical example of explotiation vs exploration refers to projects in which the course is unknown but the odds for every decision stage are known. Pychron (talk) 14:13, 12 August 2010 (UTC)