Nonadaptive Stochastic Score Classification and Explainable Half-Space Evaluation
研究非自适应随机分数分类的序贯测试问题，提出首个常数因子近似算法，并通过实验验证其成本平均不超过信息论下界的50%。
Nonadaptive Stochastic Score Classification Sequential testing problems involve a system with several components, each of which is working with some independent probability. The working/failed status of each component can be determined by performing a test, which is usually expensive. So, the goal is to perform tests in a carefully chosen sequence until the overall system status can be evaluated. These problems arise in a variety of applications, such as healthcare, manufacturing, and telecommunication. A common task in these applications is to categorize the system into one of several classes that correspond to the system status being poor, fair, good, excellent, etc. In “Nonadaptive Stochastic Score Classification and Explainable Half-Space Evaluation,” Ghuge, Gupta, and Nagarajan provide the first constant-factor approximation algorithm for this problem. Moreover, the resulting policy is nonadaptive, which results in significant savings in computational time. The authors also validate their theoretical results via computational experiments, where they observe that their algorithm’s cost is on average at most 50% more than an information-theoretic lower bound.