Optimization Algorithms#

Class Family Overview#

All optimization algorithms, regardless of their nature, are represented by classes implementing the OptimizationAlgorithm interface. Since optimization algorithms are very diverse, the OptimizationAlgorithm interface only requires a single method to be implemented: the minimize method.

Summary of Optimization Algorithm Classes#

Summary of classes in the `pybrops.opt.algo` module#
Class Name	Class Type	Class Description
`OptimizationAlgorithm`	Abstract	Interface for all optimization algorithm child classes.
`BinaryOptimizationAlgorithm`	Abstract	Interface for all optimization algorithm child classes optimizing problems in binary decision spaces.
`IntegerOptimizationAlgorithm`	Abstract	Interface for all optimization algorithm child classes optimizing problems in integer decision spaces.
`RealOptimizationAlgorithm`	Abstract	Interface for all optimization algorithm child classes optimizing problems in real decision spaces.
`SubsetOptimizationAlgorithm`	Abstract	Interface for all optimization algorithm child classes optimizing problems in subset decision spaces.
`BinaryGeneticAlgorithm`	Concrete	General genetic algorithm for optimizing single-objective problems in binary decision spaces.
`IntegerGeneticAlgorithm`	Concrete	General genetic algorithm for optimizing single-objective problems in integer decision spaces.
`RealGeneticAlgorithm`	Concrete	General genetic algorithm for optimizing single-objective problems in real decision spaces.
`SubsetGeneticAlgorithm`	Concrete	General genetic algorithm for optimizing single-objective problems in subset decision spaces.
`NSGA2BinaryGeneticAlgorithm`	Concrete	NSGA-II genetic algorithm for optimizing multi-objective problems in binary decision spaces.
`NSGA2IntegerGeneticAlgorithm`	Concrete	NSGA-II genetic algorithm for optimizing multi-objective problems in integer decision spaces.
`NSGA2RealGeneticAlgorithm`	Concrete	NSGA-II genetic algorithm for optimizing multi-objective problems in real decision spaces.
`NSGA2SubsetGeneticAlgorithm`	Concrete	NSGA-II genetic algorithm for optimizing multi-objective problems in subset decision spaces.
`SteepestDescentSubsetHillClimber`	Concrete	Steepest descent hill-climbing algorithm for optimizing single-objective problems in subset decision spaces.

Loading Optimization Algorithm Modules#

Importing optimization algorithm classes can be accomplished using the following import statements:

# import OptimizationAlgorithm classes (abstract interface classes)
from pybrops.opt.algo.OptimizationAlgorithm import OptimizationAlgorithm
from pybrops.opt.algo.BinaryOptimizationAlgorithm import BinaryOptimizationAlgorithm
from pybrops.opt.algo.IntegerOptimizationAlgorithm import IntegerOptimizationAlgorithm
from pybrops.opt.algo.RealOptimizationAlgorithm import RealOptimizationAlgorithm
from pybrops.opt.algo.SubsetOptimizationAlgorithm import SubsetOptimizationAlgorithm

# import Genetic Algorithm classes (concrete implementation classes)
from pybrops.opt.algo.BinaryGeneticAlgorithm import BinaryGeneticAlgorithm
from pybrops.opt.algo.IntegerGeneticAlgorithm import IntegerGeneticAlgorithm
from pybrops.opt.algo.RealGeneticAlgorithm import RealGeneticAlgorithm
from pybrops.opt.algo.SubsetGeneticAlgorithm import SubsetGeneticAlgorithm

# import NSGA-II Genetic Algorithm classes (concrete implementation classes)
from pybrops.opt.algo.NSGA2BinaryGeneticAlgorithm import NSGA2BinaryGeneticAlgorithm
from pybrops.opt.algo.NSGA2IntegerGeneticAlgorithm import NSGA2IntegerGeneticAlgorithm
from pybrops.opt.algo.NSGA2RealGeneticAlgorithm import NSGA2RealGeneticAlgorithm
from pybrops.opt.algo.NSGA2SubsetGeneticAlgorithm import NSGA2SubsetGeneticAlgorithm

# import steepest descent hill climber algorithm (concrete implementation class)
from pybrops.opt.algo.SteepestDescentSubsetHillClimber import SteepestDescentSubsetHillClimber

Constructing Optimization Algorithms#

Within PyBrOpS there are several pre-built optimization problems. The following code examples below illustrate the construction of example single- and multi-objective optimization algorithms.

Single-objective optimization algorithms#

The code below demonstrates the construction of a single-objective real-valued genetic algorithm.

# create a real-valued single-objective genetic algorithm
soalgo = RealGeneticAlgorithm(ngen = 100, pop_size = 100)

Multi-objective optimization algorithms#

The code below demonstrates the construction of a multi-objective real-valued NSGA-II genetic algorithm.

# create a real-valued multi-objective genetic algorithm
moalgo = NSGA2RealGeneticAlgorithm(ngen = 100, pop_size = 100)

Minimizing Optimization Problems#

Minimizing an optimization problem can be accomplished using the minimize method. Examples of how to use this method for single- and multi-objective problems are illustrated below. Intuitively, single-objective optimization algorithms should be used to optimize single-objective optimization problems, and multi-objective optimization algorithms should be used to optimize multi-objective optimization problems.

Optimization of single-objective problems#

# minimize single-objective problem and get solution
sosoln = soalgo.minimize(prob = soprob)

Optimization of multi-objective problems#

# minimize multi-objective problem and get solution
mosoln = moalgo.minimize(prob = moprob)