Surrogate-assisted Simulated Annealing

Black-box optimization refers to optimization problems where the objective function cannot be expressed in a closed form. In combinatorial optimization, black-box functions are often costly to evaluate, making traditional local search methods impractical due to their high evaluation requirements. Surrogate-assisted Optimization (SaO), which optimizes an approximation of the objective function, is a common approach. However, its effectiveness depends on the accuracy of the surrogate model. In this study, we propose a new method that combines Simulated Annealing (SA) with SaO. This method aims to balance exploration and exploitation by using a surrogate model for the energy in SA. We applied this method to several benchmark problems and compared its performance with existing SaO methods. The results show that our approach outperforms existing methods, particularly for functions that are difficult to approximate with surrogate models.