import random as rnd
import copy
from NiaPy.benchmarks.utility import Utility
__all__ = ['GeneticAlgorithm']
class Chromosome:
def __init__(self, D, LB, UB):
self.D = D
self.LB = LB
self.UB = UB
self.Solution = []
self.Fitness = float('inf')
self.generateSolution()
def generateSolution(self):
self.Solution = [self.LB + (self.UB - self.LB) * rnd.random()
for _i in range(self.D)]
def evaluate(self):
self.Fitness = Chromosome.FuncEval(self.D, self.Solution)
def repair(self):
for i in range(self.D):
if self.Solution[i] > self.UB:
self.Solution[i] = self.UB
if self.Solution[i] < self.LB:
self.Solution[i] = self.LB
def __eq__(self, other):
return self.Solution == other.Solution and self.Fitness == other.Fitness
def toString(self):
print([i for i in self.Solution])
[docs]class GeneticAlgorithm:
r"""Implementation of Genetic algorithm.
**Algorithm:** Genetic algorithm
**Date:** 2018
**Author:** Uros Mlakar
**License:** MIT
"""
def __init__(self, D, NP, nFES, Ts, Mr, gamma, benchmark):
r"""**__init__(self, D, NP, nFES, Ts, Mr, gamma, benchmark)**.
Arguments:
D {integer} -- dimension of problem
NP {integer} -- population size
nFES {integer} -- number of function evaluations
Ts {integer} -- tournament selection
Mr {decimal} -- mutation rate
gamma {decimal} -- minimum frequency
benchmark {object} -- benchmark implementation object
Raises:
TypeError -- Raised when given benchmark function which does not exists.
"""
self.benchmark = Utility().get_benchmark(benchmark)
self.NP = NP # population size; number of search agents
self.D = D # dimension of the problem
self.Ts = Ts # tournament selection
self.Mr = Mr # mutation rate
self.gamma = gamma # minimum frequency
self.Lower = self.benchmark.Lower # lower bound
self.Upper = self.benchmark.Upper # upper bound
self.Population = []
self.nFES = nFES # number of function evaluations
self.FEs = 0 # function evaluations
self.Done = False
Chromosome.FuncEval = staticmethod(self.benchmark.function())
self.Best = Chromosome(self.D, self.Lower, self.Upper)
[docs] def checkForBest(self, pChromosome):
"""Check best solution."""
if pChromosome.Fitness <= self.Best.Fitness:
self.Best = copy.deepcopy(pChromosome)
[docs] def TournamentSelection(self):
"""Tournament selection."""
indices = list(range(self.NP))
rnd.shuffle(indices)
tPop = []
for i in range(self.Ts):
tPop.append(self.Population[indices[i]])
tPop.sort(key=lambda x: x.Fitness)
self.Population.remove(tPop[0])
self.Population.remove(tPop[1])
return tPop[0], tPop[1]
[docs] def CrossOver(self, parent1, parent2):
"""Crossover."""
alpha = [-self.gamma + (1 + 2 * self.gamma) * rnd.random()
for i in range(self.D)]
child1 = Chromosome(self.D, self.Lower, self.Upper)
child2 = Chromosome(self.D, self.Lower, self.Upper)
child1.Solution = [alpha[i] * parent1.Solution[i] +
(1 - alpha[i]) * parent2.Solution[i] for i in range(self.D)]
child2.Solution = [alpha[i] * parent2.Solution[i] +
(1 - alpha[i]) * parent1.Solution[i] for i in range(self.D)]
return child1, child2
[docs] def Mutate(self, child):
"""Mutation."""
for i in range(self.D):
if rnd.random() < self.Mr:
sigma = 0.20 * float(child.UB - child.LB)
child.Solution[i] = min(
max(rnd.gauss(child.Solution[i], sigma), child.LB), child.UB)
[docs] def init(self):
"""Initialize population."""
for i in range(self.NP):
self.Population.append(Chromosome(self.D, self.Lower, self.Upper))
self.Population[i].evaluate()
self.checkForBest(self.Population[i])
[docs] def tryEval(self, c):
"""Check evaluations."""
if self.FEs < self.nFES:
self.FEs += 1
c.evaluate()
else:
self.Done = True
[docs] def run(self):
"""Run."""
self.init()
self.FEs = self.NP
while not self.Done:
for _k in range(int(self.NP / 2)):
parent1, parent2 = self.TournamentSelection()
child1, child2 = self.CrossOver(parent1, parent2)
self.Mutate(child1)
self.Mutate(child2)
child1.repair()
child2.repair()
self.tryEval(child1)
self.tryEval(child2)
tPop = [parent1, parent2, child1, child2]
tPop.sort(key=lambda x: x.Fitness)
self.Population.append(tPop[0])
self.Population.append(tPop[1])
for i in range(self.NP):
self.checkForBest(self.Population[i])
return self.Best.Fitness