surrogate_based module¶

A class for surrogate-based optimization.

class gemseo_mlearning.algos.opt.core.surrogate_based.SurrogateBasedOptimizer(problem, acquisition_algorithm, doe_size=0, doe_algorithm='OT_OPT_LHS', doe_options=mappingproxy({}), regression_algorithm='GaussianProcessRegressor', regression_options=mappingproxy({}), acquisition_options=mappingproxy({}))[source]¶

Bases: object

An optimizer based on surrogate models.

Parameters:

acquisition_algorithm (str) – The name of the algorithm to optimize the data acquisition criterion. N.B. this algorithm must handle integers if some of the optimization variables are integers.
problem (OptimizationProblem) – The optimization problem.
doe_size (int) –
The size of the initial DOE. Should be 0 if the DOE algorithm does not have a n_samples option.

By default it is set to 0.
doe_algorithm (str) –
The name of the algorithm for the initial sampling.

By default it is set to “OT_OPT_LHS”.
doe_options (Mapping[str, DOELibraryOptionType]) –
The options of the algorithm for the initial sampling.

By default it is set to {}.
regression_algorithm (str) –
The name of the regression algorithm for the objective function.

By default it is set to “GaussianProcessRegressor”.
regression_options (Mapping[str, MLAlgoParameterType]) –
The options of the regression algorithm for the objective function.

By default it is set to {}.
acquisition_options (Mapping[str, OptimizationLibraryOptionType]) –
The options of the algorithm to optimize the data acquisition criterion.

By default it is set to {}.

execute(number_of_acquisitions)[source]¶

Execute the surrogate-based optimization.

Parameters:: number_of_acquisitions (int) – The number of learning points to be acquired.
Returns:: The termination message.
Return type:: str