An Exploitation-Enhanced Bayesian Optimization Algorithm for High-Dimensional Expensive Problems

The Bayesian optimization (BO) algorithm is widely used to solve expensive optimization problems. However, when dealing with high-dimensional problems, the accuracy of the global Gaussian process (GP) model is often inadequate due to the limited number of training points. As a result, the search based on the expected improvement criterion can lead to misguided exploration. To address this issue, we propose an exploitation-enhanced Bayesian optimization (EE-BO) algorithm. Our approach incorporates a local GP model built around the evaluated solution from the previous iteration, which is used to find the next infill solution if the current selection from the global GP model is not an improvement. The inclusion of the local model mitigates the impact of inaccurate models and enhances the algorithm's ability to perform local searches when the global model struggles to find better solutions. Our numerical experiments show that the proposed EE-BO algorithm outperforms the vanilla BO algorithm and achieves competitive performance compared to five state-of-the-art algorithms.