Surrogate modeling#
Cross-layer
TrainingSet builds training data from an Archive or Population, and Surrogate predicts objective values from that data.
SurrogateManager coordinates when fitting and prediction occur, while AcquisitionFunction converts predictions into scores for candidate selection.
AccuracyBasedSurrogateSwitcher switches the SurrogateManager or OptimizationStrategy according to prediction-accuracy evaluations.
Surrogate
Fits a model from training data and returns predictions for candidates.
SurrogateManager
Coordinates model fitting and batch prediction.
TrainingSet
Builds training inputs and labels from an Archive or Population.
AcquisitionFunction
Converts predictions into scalar scores for candidate selection.
AccuracyBasedSurrogateSwitcher
Evaluates Surrogate accuracy and switches the configuration used during a run.