mach_opt package

mach_opt module

Module holding classes required for design optimization.

This module holds the classes required for optimizing a design using pygmo in MachEval.

class mach_opt.mach_opt.DataHandler(archive_filepath, designer_filepath)

Bases: object

Parent class for data handlers

get_archive_data()

get_pareto_data(): Return data of Pareto optimal designs

get_pareto_fitness_freevars(): Extract fitness and free variables for Pareto optimal designs

load_from_archive(): Load data from Pickle optimization archive

load_object(filename): load object from specified filename

save_designer(designer): Save designer used in optimization

save_object(obj, filename): Save object to specified filename

save_to_archive(x, design, full_results, objs)

Save machine evaluation data to optimization archive using Pickle

Parameters:

x – Free variables used to create design
design – Created design
full_results – Input, output, and results corresponding to each step of an evaluator
objs – Fitness values corresponding to a design

class mach_opt.mach_opt.Design

Bases: ABC

Parent class for all designs

class mach_opt.mach_opt.DesignOptimizationMOEAD(design_problem)

Bases: object

initial_pop(pop_size)

load_pop(filepath, pop_size)

run_optimization(pop, gen_size, filepath=None)

save_pop(filepath, pop)

class mach_opt.mach_opt.DesignProblem(designer: Designer, evaluator: Evaluator, design_space: DesignSpace, dh: DataHandler, invalid_design_objs=None)

Bases: object

Class to create, evaluate, and optimize designs

designer: Objects which convert free variables to a design.

evaluator: Objects which evaluate the performance of different designs.

design_space: Objects which characterizes the design space of the optimization.

dh: Data handlers which enable saving optimization results and its resumption.

invalid_design_objs: List of (large) objective values to use for invalid designs

fitness(x: tuple) → tuple

Calculates the fitness or objectives of each design based on evaluation results.

This function creates, evaluates, and calculates the fitness of each design generated by the optimization algorithm. It also saves the results and handles invalid designs.

Parameters:: x – The list of free variables required to create a complete design
Returns:: Returns the fitness of each design
Return type:: objs
Raises:: e – The errors encountered during design creation or evaluation apart from the InvalidDesign error

get_bounds(): Returns bounds for optimization problem

get_nobj(): Returns number of objectives of optimization problem

class mach_opt.mach_opt.DesignSpace(*args, **kwargs)

Bases: Protocol

Parent class for a optimization DesignSpace classes

abstract bounds() → tuple

abstract check_constraints(full_results) → bool

abstract get_objectives(valid_constraints, full_results) → tuple

abstract n_obj() → int

class mach_opt.mach_opt.Designer(*args, **kwargs)

Bases: Protocol

Parent class for all designers

abstract create_design(x: tuple) → Design

class mach_opt.mach_opt.Evaluator(*args, **kwargs)

Bases: Protocol

Parent class for all design evaluators

abstract evaluate(design: Design) → Any

exception mach_opt.mach_opt.InvalidDesign(message='Invalid Design')

Bases: Exception

Exception raised for invalid designs

class mach_opt.mach_opt.OptiData(x, design, full_results, objs)

Bases: object

Object template for serializing optimization results with Pickle