Abstract:Optimizing an expensive, black-box function <math>f(·)</math> is challenging when its input space is high-dimensional. Sequential design frameworks first model <math>f(·)</math> with a surrogate function and then optimize an acquisition function to determine input settings to evaluate next. Optimization of both <math>f(·)</math> and the acquisition function benefit from effective dimension reduction. Global variable selection detects and removes input variables that do not affect <math>f(·)</math> across the input space. Further dimension reduction may be possible if we consider local variable selection around the current optimum estimate. We develop a sequential design algorithm called sequential optimization in locally important dimensions (SOLID) that incorporates global and local variable selection to optimize a continuous, differentiable function. SOLID performs local variable selection by comparing the surrogate's predictions in a localized region around the estimated optimum with the p alternative predictions made by removing each input variable. The search space of the acquisition function is further restricted to focus only on the variables that are deemed locally active, leading to greater emphasis on refining the surrogate model in locally active dimensions. A simulation study across multiple test functions and an application to the Sarcos robot dataset show that SOLID outperforms conventional approaches. <a class="ext-link" href="https://doi.org/10.1080/00401706.2020.1714738">Supplementary materials</a> for this article are available online.

A sequential procedure for neighborhood selection-of-the-best in optimization via simulation

Fully Sequential Indifference‐zone Selection Procedures with Variance‐dependent Sampling

Indifference-Zone-Free Selection of the Best

Several Algorithms of Global Optimal Search

An Indifference-Zone Selection Procedure with Minimum Switching and Sequential Sampling

A New Framework of Designing Sequential Ranking-and-selection Procedures.

Fully Sequential Ranking and Selection Procedures with PAC Guarantee.

Selection of the Best in the Presence of Subjective Stochastic Constraints

Sequential Optimization in Locally Important Dimensions

Optimization via Rejection-Free Partial Neighbor Search

Sequential Design for Optimal Stopping Problems

Discrete optimization via simulation: algorithms and error control

A novel batch-selection strategy for parallel global optimization

Discrete Optimization Via Simulation Using Coordinate Search

Chapter 2 Discrete Optimization via Simulation

Adaptive Sequential Procedures for Quantile Selection with Importance Sampling

A frequentist selection-of-the-best procedure without indifference zone

Selecting the Best Simulated Design with the Expected Opportunity Cost Bound.

Non-Myopic Knowledge Gradient Policy for Ranking and Selection.

A Sequential Sampling Procedure for Stochastic Programming

A Non-Parametric Sequential Procedure for the Generalized Partition Problem