Analysis‐of‐Marginal‐Tail‐Means: A Robust Method for Discrete Black‐Box Optimization
Simon Mak, C. F. Jeff Wu · Wiley StatsRef: Statistics Reference Online · 2021
Abstract We introduce a robust optimization method, called the Analysis‐of‐marginal‐Tail‐Means (ATM), which was proposed for discrete black‐box optimization. ATM has applications in a variety of engineering problems (e.g., manufacturing optimization and product design), where the objective function is black box, and the design space is discrete. The key novelty in ATM is a data‐tuned compromise between rank‐ and model‐based optimization via marginal tail means, which allows it to adaptively exploit marginal structure for robust optimization. We demonstrate the effectiveness of ATM in two real‐world problems: the first on robust parameter design of a circular piston and the second on product family design of a thermistor network.