optimizer

Module Contents

Classes

Functions

Attributes

default

class OptimizerSettings(theDevice, theObjectives, theConstraints, theOptimizationVariables, theOptimizationAlgorithm=None, theMathsToPhysics=None, theCharacterization=None)

Bases: optimeed.core.SaveableObject

Abstract class for dynamically type-hinted objects. This class is to solve the special case where the exact type of an attribute is not known before runtime, yet has to be saved.

get_additional_attributes_to_save()

Return list of attributes corresponding to object, whose type cannot be determined statically (e.g. topology change)

get_additional_attributes_to_save_list()

Same behavior as get_additional_attributes_to_save, but where the attributes contains list of unknown items

get_device()

get_M2P()

get_charac()

get_optivariables()

get_objectives()

get_constraints()

get_optialgorithm()

class _Evaluator(optimization_parameters: OptimizerSettings)

This is the main class that serves as evaluator. This class is NOT process safe (i.e., copy of it might be generated upon process call)

start()

evaluate(x)

Evaluates the performances of device associated to entrance vector x. Outputs the objective function and the constraints, and other data used in optiHistoric.

This function is NOT process safe: “self.” is a FORK in multiprocessing algorithms. It means that the motor originally contained in self. is modified only in the fork, and only gathered by reaching the end of the fork.

Parameters:x – Input mathematical vector from optimization algorithm Returns:dictionary, containing objective values (list of scalar), constraint values (list of scalar), and other info (motor, time)

reevaluate_solutions(x_solutions)

class _ArrayEvaluator(optimization_parameters: OptimizerSettings)

Bases: _Evaluator

Same as _evaluator, using array as inputs. Allows to evaluate all generation at once, in case of numpy array computations. Drawbacks: devices can not be visualized and saved. Use this with care.

evaluate(list_of_x)

Evaluates the performances of device associated to entrance vector x. Outputs the objective function and the constraints, and other data used in optiHistoric.

This function is NOT process safe: “self.” is a FORK in multiprocessing algorithms. It means that the motor originally contained in self. is modified only in the fork, and only gathered by reaching the end of the fork.

Parameters:x – Input mathematical vector from optimization algorithm Returns:dictionary, containing objective values (list of scalar), constraint values (list of scalar), and other info (motor, time)

reevaluate_solutions(x_solutions)

run_optimization(optimizer_settings: OptimizerSettings, opti_historic, max_opti_time_sec=10, return_x_solutions=False, array_evaluator=False, initialVectorGuess=None)

Perform the optimization.

Parameters:

• optimizer_settings – OptimizerSettings containing all information in models, objectives and optimization variable
• opti_historic – OptiHistoric to log evaluations and results
• max_opti_time_sec – Maximum optimization time (default stopping criterion, unless modified in algorithm)
• return_x_solutions – If True, returns raw parameters in reults
• array_evaluator – If True, evaluate each generation at once using numpy array. Use it only with care, as it dismisses some features (expert mode)

Returns:

list of the best optimized devices, convergence information and [if return_x_solutions=True] best solutions