论文标题
动态库存问题的无递归功能近似
A recursion-free functional approximation for the dynamic inventory problem
论文作者
论文摘要
我们考虑了非平稳需求的动态库存问题。长期以来，已经知道非平稳（S，S）策略对于此问题是最佳的。但是，找到最佳策略参数仍然是一个计算挑战，因为它需要解决大规模的随机动态程序。为了解决这个问题，我们为问题的最佳成本函数设计了无递归近似值。这使我们能够以启发性计算策略参数，而无需诉诸随机动态程序。该启发式是易于理解和使用的，因为它遵循凸的基本方法最小化和最短的路径，但它非常有效，并且较早的启发式方法表现出色。
We consider the dynamic inventory problem with non-stationary demands. It has long been known that non-stationary (s, S) policies are optimal for this problem. However, finding optimal policy parameters remains a computational challenge as it requires solving a large-scale stochastic dynamic program. To address this, we devise a recursion-free approximation for the optimal cost function of the problem. This enables us to compute policy parameters heuristically, without resorting to a stochastic dynamic program. The heuristic is easy-to-understand and -use since it follows by elementary methods of convex minimization and shortest paths, yet it is very effective and outperforms earlier heuristics.
