Typically, a set of evaluated genotypes and their real fitness
scores are used to perform the supervised training of
an MLP-based artificial neural network; e.g., [57]. However,
other approaches have been explored, e.g., Kriging [58],
clustering [55], support vector regression [59], radial-basis
functions [60], and sequential parameter optimization [61].
The surrogate model is subsequently used to compute estimated
fitness values for the EA to utilize. The model must be
periodically retrained with new individuals under a controlled
evolutionary approach to prevent convergence on local optima.
Retraining can be performed by taking either an individual or
generational approach [62]. In the individual approach, n number
of individuals in the population are chosen and evaluated
with the real fitness function each generation. In the generational
approach, the entire population is evaluated on the real
fitness function each n-th generation. Resampling methods and
surrogate model validation remain an important and ongoing
area of research, enabling the comparison and optimization of
models [63]. Both global modeling and local modeling using
trust regions (e.g., samples within a certain Euclidean distance)
are popular approaches