This paper considers the maximization of the expected maximum value of a portfolio of random variables subject to a budget constraint. We refer to this as the optimal college application problem. When each variable's cost, or each college's application fee, is identical, we show that the optimal portfolios are nested in the budget constraint, yielding an exact polynomial-time algorithm. When colleges differ in their application fees, we show that the problem is NP-complete. We provide four algorithms for this more general setup: a branch-and-bound routine, a dynamic program that produces an exact solution in pseudopolynomial time, a different dynamic program that yields a fully polynomial-time approximation scheme, and a simulated-annealing heuristic. Numerical experiments demonstrate the algorithms' accuracy and efficiency.
翻译:本文考虑了受预算制约的随机变量组合的预期最大值最大化。 我们将此称为最佳大学应用问题。 当每个变量的成本或每个学院的应用费相同时, 我们显示最佳组合在预算约束下嵌套, 产生一个精确的多元时间算法。 当各学院在应用费上存在差异时, 我们显示问题是NP- 完整的。 我们为这个更普遍的设置提供了四种算法: 一个分支和约束常规, 一个动态程序, 在假球时代产生一个精确的解决方案, 一个不同的动态程序, 产生一个完全的多元时间近似计划, 以及一个模拟的静脉冲。 数值实验显示了算法的准确性和效率 。