5 ms·

Doesn't this problem basically take the Feynman restaurant problem and replaces meal with woman? http://www.feynmanlectures.info/exercises/Feynmans_restaurant_

• SlyShy 16y ago
  In short, no. You are calculating the expected value differently in the two problems.
• presidentender 16y ago
  The difference in this case is that you can't switch back to a woman you've already "sampled." With Feynman's restaurant, we can sample ten dishes, and choose the fourth dish for the rest of eternity. With the wife selection algorithm (a modification of the "Sultan's Daughter" problem) we're only allowed to marry the current candidate, or take a pass on her. Exes are off-limits.
• secretasiandan 16y ago
  I believe this problem is a lot closer to the Secretary Problem http://en.wikipedia.org/wiki/Secretary_problem http://en.wikipedia.org/wiki/Secretary_problem As he notes however, he does not try to choose "the very best candidate", for which the optimal solution is to pass on the first 1/e candidates. Instead, he tries to maximize the "goodness" of his wife selection. In short : maximum likelihood soultion is to pass on (1/e)*N, minimum mean squared error solution is to pass on sqrt(N). This of course assumes you can estimate N well.