Skip to content

Default Function Estimator Plugin

ropt.plugins.function_estimator.default.DefaultFunctionEstimator

Bases: FunctionEstimator

The default implementation for function estimation strategies.

This class provides methods for combining objective function values and gradients from an ensemble of realizations into a single representative value or gradient. The specific method is configured via the FunctionEstimatorConfig in the main EnOptConfig.

Supported Methods:

  • mean (or default): Calculates the combined function value as the weighted mean of the individual realization function values. The combined gradient is calculated as the weighted mean of the individual realization gradients (unless merge_realizations is true, in which case the pre-merged gradient is used directly).

  • stddev: Calculates the combined function value as the weighted standard deviation of the individual realization function values. The combined gradient is calculated using the chain rule based on the standard deviation formula. This method requires at least two realizations with non-zero weights and is incompatible with merge_realizations=True for gradient calculation.