surrogateassisted
Surrogateassisted, also written as surrogate-assisted, refers to a class of optimization methods that combine a computationally inexpensive surrogate model with an outer search strategy to optimize an expensive objective function. The surrogate provides fast approximations of the true objective, enabling more iterations of evaluation at a reduced computational cost. This approach is widely used in engineering design, materials discovery, hyperparameter tuning, and other areas where each evaluation of the objective is costly or time-consuming.
Common surrogate models include Gaussian processes (kriging), radial basis function networks, polynomial regression, support vector machines,
Advantages include reduced number of expensive evaluations and the ability to incorporate uncertainty. Limitations include potential