Provably Near-Optimal Sampling-Based Policies for Stochastic Inventory Control Models

摘要: In this paper, we consider two fundamental inventory models, the single-period newsvendor problem and its multiperiod extension, but under assumption that explicit demand distributions are not known only information available is a set of independent samples drawn from true distributions. Under given explicitly, these models well studied relatively straightforward to solve. However, in most real-life scenarios, available, or they too complex work with. Thus, sampling-driven algorithmic framework very attractive, both practice theory. We shall describe how compute sampling-based policies, is, policies computed based on observed demands without any access to, assumptions on, Moreover, establish bounds number required guarantee that, with high probability, expected cost arbitrarily close (i.e., small relative error) compared optimal which have full The develop general, easy compute, do depend at all specific