doe_scenario module¶
Scenario which drivers are Design of Experiments¶
-
class
gemseo.core.doe_scenario.
DOEScenario
(disciplines, formulation, objective_name, design_space, name=None, **formulation_options)[source]¶ Bases:
gemseo.core.scenario.Scenario
Design of Experiments scenario, based on MDO scenario but with a DOE driver.
The main differences between Scenario and MDOScenario are the allowed inputs in the MDOScenario.json, which differs from DOEScenario.json, at least on the driver names
MDO Problem description: links the disciplines and the formulation to create an optimization problem. Use the class by instantiation. Create your disciplines beforehand.
Specify the formulation by giving the class name such as the string “MDF”
The reference_input_data is the typical input data dict that is provided to the run method of the disciplines
Specify the objective function name, which must be an output of a discipline of the scenario, with the “objective_name” attribute
If you want to add additional design constraints, use the add_user_defined_constraint method
To view the results, use the “post_process” method after execution. You can view:
the design variables history, the objective value, the constraints, by using: scenario.post_process(“OptHistoryView”, show=False, save=True)
Quadratic approximations of the functions close to the optimum, when using gradient based algorithms, by using: scenario.post_process(“QuadApprox”, method=”SR1”, show=False, save=True, function=”my_objective_name”, file_path=”appl_dir”)
Self Organizing Maps of the design space, by using: scenario.post_process(“SOM”, save=True, file_path=”appl_dir”)
To list post processings on your setup, use the method
Scenario.posts
. For more details on their options, go to the gemseo.post package.Constructor, initializes the DOE scenario Objects instantiation and checks are made before run intentionally
- Parameters
disciplines – the disciplines of the scenario
formulation – the formulation name, the class name of the formulation in gemseo.formulations
objective_name – the objective function name
design_space – the design space
name – scenario name
formulation_options – options for creation of the formulation
-
EVAL_JAC
= 'eval_jac'¶
-
N_SAMPLES
= 'n_samples'¶