Prophet Inequalities for Independent and Identically Distributed Random Variables from an Unknown Distribution

• 成果类型: Article
• 署名作者: Correa, Jose; Dutting, Paul; Fischer, Felix; Schewior, Kevin
• 署名单位: Universidad de Chile; University of London; London School Economics & Political Science; University of London; Queen Mary University London; University of Cologne
• 刊物名称: MATHEMATICS OF OPERATIONS RESEARCH
• ISSN/ISSBN: 0364-765X
• DOI: 10.1287/moor.2021.1167
• 发表日期: 2022
• 页码: 1287-1309
• 关键词: supremum expectations; stop rule; maximum

摘要

A central object of study in optimal stopping theory is the single-choice prophet inequality for independent and identically distributed random variables. given a sequence of random variables X-1, . . . , X-n drawn independently from the same distribution, the goal is to choose a stopping time tau such that for the maximum value of alpha and for all distributions, E[X-tau] >= alpha center dot E[max X-t(t)]. What makes this problem challenging is that the decision whether tau = t may only depend on the values of the random variables X-1, . . . , X-t and on the distribution F. For a long time, the best known bound for the problem had been alpha >= 1 - 1/e approximate to 0.632, but recently a tight bound of alpha approximate to 0.745 was obtained. The case where F is unknown, such that the decision whether tau = t may depend only on the values of the random variables X-1, . . . , X-t, is equally well motivated but has received much less attention. A straightforward guarantee for this case of alpha >= 1/e approximate to 0.368 can be derived from the well-known optimal solution to the secretary problem, where an arbitrary set of values arrive in random order and the goal is to maximize the probability of selecting the largest value. We show that this bound is in fact tight. We then investigate the case where the stopping time may additionally depend on a limited number of samples from F, andwe show that, even with o(n) samples, alpha <= 1/e. On the other hand, n samples allow for a significant improvement, whereas O(n(2)) samples are equivalent to knowledge of the distribution. specifically, with n samples, alpha >= 1 - 1/e approximate to 0.632 and alpha <= ln (2) approximate to 0.693, and with O(n(2)) samples, alpha >= 0.745 - epsilon for any epsilon > 0.