Efficient global optimization aims to identify the global optimum of an expensive black-box function within a limited evaluation budget. Gaussian process Thompson sampling (GP-TS) is a prominent method for this task, but its theoretical convergence behavior has remained an active area of study.

The paper, announced on arXiv, investigates the convergence rates of GP-TS under a limited evaluation budget. The abstract does not disclose specific results, but the analysis targets the core question of how quickly the method can approach the global optimum as the number of evaluations grows.

This work contributes to the theoretical foundation of Bayesian optimization, offering insight into when GP-TS can be expected to perform reliably in practical settings where function evaluations are costly.