Pandora's Box is a central problem in decision making under uncertainty that can model various real life scenarios. In this problem we are given $n$ boxes, each with a fixed opening cost, and an unknown value drawn from a known distribution, only revealed if we pay the opening cost. Our goal is to find a strategy for opening boxes to minimize the sum of the value selected and the opening cost paid. In this work we revisit Pandora's Box when the value distributions are correlated, first studied in Chawla et al. (arXiv:1911.01632). We show that the optimal algorithm for the independent case, given by Weitzman's rule, directly works for the correlated case. In fact, our algorithm results in significantly improved approximation guarantees compared to the previous work, while also being substantially simpler. We finally show how to implement the rule given only sample access to the correlated distribution of values. Specifically, we find that a number of samples that is polynomial in the number of boxes is sufficient for the algorithm to work.
翻译:潘多拉的“ 盒子” 是一个在不确定性下决策的中心问题, 它可以模拟各种真实生活情景。 在这个问题中, 我们得到的是一美元盒子, 每个都有固定的开场成本, 以及从已知发行中提取的未知值, 只有在我们支付开场成本时才透露出来。 我们的目标是找到一个策略, 打开盒子, 以尽量减少所选价值和所支付的开场成本的总和。 在这项工作中, 当价值分布是相互关联的时, 我们重新查看了潘多拉的“ 盒子 ”, 最初在Chawla 等人( arXiv: 1911. 01632) 中研究过, 我们发现, 由魏茨曼 规则给出的独立案例的最佳算法, 直接为相关案例运作。 事实上, 我们的算法结果大大改进了近似保证, 同时也大大简化了。 我们最后展示了如何执行这一规则, 只给相关价值分布的样本使用。 具体地说, 我们发现, 在框数中具有多数值的样本, 足以使算法发挥作用 。