def __init__(self, representation, size, fitness_func, selection_func,
crossover_func, mutation_func, natural_fitness,
crossover_probability, mutation_probability, elite_count,
tournament_size):
"""
Constructor
:param representation: the representation dictionary for each
genome in this population
:param size: the number of genomes in this population
:param fitness_func: the user-specified fitness function
:param selection_func: the user-specified selection function
:param crossover_func: the user-specified crossover function
:param mutation_func: the user-specified mutation function
:param natural_fitness: use natural fitness values, i.e. higher
fitness value implies fitter individual
:param crossover_probability: the user-specified crossover probability
:param mutation_probability: the user-specified mutation probability
:param elite_count: the number of fittest individuals to
exclude from selection/crossover/mutation
:param tournament_size: the size of a tournament selection group
"""
self.population = list()
self.elites = list()
#-----------------------------------------------------------------------
# Ensure even population size
#-----------------------------------------------------------------------
if size % 2 != 0:
size += 1
self.size = size
#-----------------------------------------------------------------------
# Ensure 0 < probability < 1
#-----------------------------------------------------------------------
if not (0 < float(crossover_probability) < 1) \
or not (0 < float(mutation_probability) < 1):
raise ValueError('probabilities must be between 0 and 1')
self.selection_func = selection_func
self.crossover_func = crossover_func
self.mutation_func = mutation_func
self.fitness_func = fitness_func
self.representation = representation
self.natural_fitness = natural_fitness
self.crossover_probability = crossover_probability
self.mutation_probability = mutation_probability
#-------------------------------------------------------------------
# Ensure even elite count
#-------------------------------------------------------------------
if elite_count % 2 != 0:
elite_count += 1
self.elite_count = elite_count
self.tournament_size = tournament_size
self.plotter = Plotter()
self.average_sigmas = list()
class Population(object):
"""
Class which represents an entire population of individuals.
"""
def __init__(self, representation, size, fitness_func, selection_func,
crossover_func, mutation_func, natural_fitness,
crossover_probability, mutation_probability, elite_count,
tournament_size):
"""
Constructor
:param representation: the representation dictionary for each
genome in this population
:param size: the number of genomes in this population
:param fitness_func: the user-specified fitness function
:param selection_func: the user-specified selection function
:param crossover_func: the user-specified crossover function
:param mutation_func: the user-specified mutation function
:param natural_fitness: use natural fitness values, i.e. higher
fitness value implies fitter individual
:param crossover_probability: the user-specified crossover probability
:param mutation_probability: the user-specified mutation probability
:param elite_count: the number of fittest individuals to
exclude from selection/crossover/mutation
:param tournament_size: the size of a tournament selection group
"""
self.population = list()
self.elites = list()
#-----------------------------------------------------------------------
# Ensure even population size
#-----------------------------------------------------------------------
if size % 2 != 0:
size += 1
self.size = size
#-----------------------------------------------------------------------
# Ensure 0 < probability < 1
#-----------------------------------------------------------------------
if not (0 < float(crossover_probability) < 1) \
or not (0 < float(mutation_probability) < 1):
raise ValueError('probabilities must be between 0 and 1')
self.selection_func = selection_func
self.crossover_func = crossover_func
self.mutation_func = mutation_func
self.fitness_func = fitness_func
self.representation = representation
self.natural_fitness = natural_fitness
self.crossover_probability = crossover_probability
self.mutation_probability = mutation_probability
#-------------------------------------------------------------------
# Ensure even elite count
#-------------------------------------------------------------------
if elite_count % 2 != 0:
elite_count += 1
self.elite_count = elite_count
self.tournament_size = tournament_size
self.plotter = Plotter()
self.average_sigmas = list()
def run(self, generations):
"""
Apply selection, crossover and mutation on the given population as many
times as the given number of generations.
:param generations: the number of generations/cycles to perform
"""
self.calculate_fitnesses()
#-----------------------------------------------------------------------
# Print a short summary
#-----------------------------------------------------------------------
print 'population size=%d, representation=%s, ' \
'crossover probability=%f, mutation probability=%f, ' \
'elite count=%d' \
% (len(self.population), self.representation,
self.crossover_probability, self.mutation_probability,
self.elite_count)
print 'selection scheme=%s, crossover scheme=%s, mutation scheme=%s, ' \
'fitness function=%s, natural_fitness=%s' \
% (self.selection_func, self.crossover_func,
self.mutation_func,
self.fitness_func, self.natural_fitness)
print 'generation=0, total fitness=%d, mean fitness=%s, ' \
'min individual=%s (len=%d), max individual=%s (len=%d)' \
% (self.total_fitness(), self.mean_fitness(),
self.min_individual().fitness(),
len(self.min_individual()),
self.max_individual().fitness(),
len(self.max_individual()))
self.plotter.update(self.mean_fitness(),
self.max_individual().fitness(),
self.min_individual().fitness())
for i in self.population:
if hasattr(i, 'average_sigmas') and i.average_sigmas is not None:
self.average_sigmas.append(sum(i.average_sigmas)
/ len(i.average_sigmas))
# print self.average_sigmas
# print len(self.average_sigmas)
#-----------------------------------------------------------------------
# Loop for each generation
#-----------------------------------------------------------------------
for i in xrange(1, generations):
#-------------------------------------------------------------------
# Perform elitism
#-------------------------------------------------------------------
self.store_elites()
#-------------------------------------------------------------------
# Select the mating pool
#-------------------------------------------------------------------
self.select_parents()
#-------------------------------------------------------------------
# Apply crossover
#-------------------------------------------------------------------
self.crossover(self.crossover_probability)
#-------------------------------------------------------------------
# Apply mutation
#-------------------------------------------------------------------
self.mutate(self.mutation_probability)
#-------------------------------------------------------------------
# Re-add the elites to the population
#-------------------------------------------------------------------
self.load_elites()
#-------------------------------------------------------------------
# Recalculate fitnesses
#-------------------------------------------------------------------
self.calculate_fitnesses()
min_individual = self.min_individual()
max_individual = self.max_individual()
# print 'generation=%d, total fitness=%d, mean fitness=%s, ' \
# 'min individual=%s (%s), max individual=%s (%s)' \
# % (i, self.total_fitness(), self.mean_fitness(),
# min_individual.genes,
# min_individual.raw_fitness(),
# max_individual.genes,
# max_individual.raw_fitness())
print 'generation=%d, total fitness=%d, mean fitness=%s, ' \
'min individual=%s (len=%d), max individual=%s (len=%d)' \
% (i, self.total_fitness(), self.mean_fitness(),
self.min_individual().fitness(),
len(self.min_individual()),
self.max_individual().fitness(),
len(self.max_individual()))
self.plotter.update(self.mean_fitness(),
max_individual.fitness(),
min_individual.fitness())
for i in self.population:
if hasattr(i, 'average_sigmas') and i.average_sigmas is not None:
self.average_sigmas.append(sum(i.average_sigmas)
/ len(i.average_sigmas))
# print self.average_sigmas
# print len(self.average_sigmas)
def __iter__(self):
return iter(self.population)
def __len__(self):
return len(self.population)
def __getitem__(self, item):
return self.population[item]
def gen_population(self):
"""
Generate an initial, random population based on the given representation
dictionary.
"""
#-----------------------------------------------------------------------
# Generate binary population
#-----------------------------------------------------------------------
if self.representation.type == 'binary':
for _ in xrange(self.size):
fmt = '{0:0' + str(self.representation.length) + 'b}'
gene = Genome(fmt.format(
random.randint(0, 2 ** self.representation.length)),
representation=self.representation,
fitness_func=self.fitness_func,
natural_fitness=self.natural_fitness)
self.population.append(gene)
#-----------------------------------------------------------------------
# Generate float value population
#-----------------------------------------------------------------------
elif self.representation.type == 'float':
for _ in xrange(self.size):
gene = Genome([random.uniform(self.representation.min,
self.representation.max)
for _ in xrange(self.representation.length)],
representation=self.representation,
fitness_func=self.fitness_func,
natural_fitness=self.natural_fitness)
self.population.append(gene)
#-----------------------------------------------------------------------
# Generate integer value population
#-----------------------------------------------------------------------
elif self.representation.type == 'int':
for _ in xrange(self.size):
gene = Genome([random.randint(self.representation.min,
self.representation.max)
for _ in xrange(self.representation.length)],
representation=self.representation,
fitness_func=self.fitness_func,
natural_fitness=self.natural_fitness)
self.population.append(gene)
#-----------------------------------------------------------------------
# Generate fixed value population
#-----------------------------------------------------------------------
elif self.representation.type == 'enum':
for _ in xrange(self.size):
#---------------------------------------------------------------
# Allow duplicates
#---------------------------------------------------------------
if self.representation.duplicates:
gene = [self.representation.values[random.randint(0,
len(self.representation.values)) - 1]
for _ in xrange(self.representation.length)]
#---------------------------------------------------------------
# Disallow duplicates
#---------------------------------------------------------------
else:
import copy
gene = copy.copy(self.representation.values)
random.shuffle(gene)
gene = Genome(gene,
representation=self.representation,
fitness_func=self.fitness_func,
natural_fitness=self.natural_fitness)
self.population.append(gene)
def calculate_fitnesses(self):
""""""
for i in self.population:
i.fitness(recalculate=True)
def update_population(self, population):
"""
Change the existing population to the new one
:param population: the new population to use
"""
self.population = population
def total_fitness(self):
"""
Return the total fitness of all of the individuals in the population
"""
return sum([i.fitness() for i in self.population])
def mean_fitness(self):
"""
Return the mean fitness of all of the individuals in the population
"""
return self.total_fitness() / len(self.population)
def max_individual(self):
"""
Return the individual with the maximum (highest) fitness of all the
individuals in the population
"""
max_indiv = self.population[0]
for i in self.population:
if i.fitness() > max_indiv.fitness():
max_indiv = i
return max_indiv
def min_individual(self):
"""
Return the individual with the minimum (lowest) fitness of all the
individuals in the population
"""
min_indiv = self.population[0]
for i in self.population:
if i.fitness() < min_indiv.fitness():
min_indiv = i
return min_indiv
def store_elites(self):
"""
Perform elitism by withholding a certain number of the fittest
individuals from selection/crossover/mutation.
"""
del self.elites[:]
for i in xrange(self.elite_count):
max_indiv = self.max_individual()
self.elites.append(max_indiv)
self.population.remove(max_indiv)
def load_elites(self):
"""
Re-add the withheld elites back into the population
"""
self.population += self.elites
def select_parents(self):
"""
Perform population selection using the user-supplied selection function.
"""
selected_parents = self.selection_func(self)
self.update_population([self.make_copy(i) for i in selected_parents])
def crossover(self, probability):
"""
Perform crossover using the user-supplied crossover function
:param probability: the probability that crossover will occur for each
pair of individuals
"""
result = list()
#-----------------------------------------------------------------------
# Loop the population in twos
#-----------------------------------------------------------------------
for male, female in self.pairwise(self.population):
child1, child2 = self.make_copy(male), self.make_copy(female)
#-------------------------------------------------------------------
# Maybe do the crossover... maybe not
#-------------------------------------------------------------------
if random.random() <= probability:
child1.genes, child2.genes = self.crossover_func(child1.genes,
child2.genes)
# male.genes = child1[:]
# female.genes = child2[:]
result.append(child1)
result.append(child2)
# result.append(Genome(child1, self.representation,
# self.fitness_func, self.natural_fitness))
# result.append(Genome(child2, self.representation,
# self.fitness_func, self.natural_fitness))
assert len(result) == len(self.population)
#-----------------------------------------------------------------------
# Switch to the new population
#-----------------------------------------------------------------------
self.update_population(result)
def mutate(self, probability):
"""
Perform mutation using the user-supplied mutation function
:param probability: the probability that mutation will occur for each
individual
"""
result = list()
for i in self.population:
#-------------------------------------------------------------------
# Make a copy of the genes
#-------------------------------------------------------------------
# i.genes = self.mutation_func(copy.deepcopy(i.genes), probability)
indiv_copy = self.make_copy(i)
genes = self.mutation_func(indiv_copy, probability)
indiv_copy.genes = genes
result.append(indiv_copy)
assert len(result) == len(self.population)
self.update_population(result)
def make_copy(self, individual):
if isinstance(individual.genes[0], Gene):
genes_copy = list()
for gene in individual.genes:
gene_copy = Gene(gene.alleles[:])
gene_copy.class_label = gene.class_label
if gene.mutation_step_sizes is not None:
gene_copy.mutation_step_sizes = gene.mutation_step_sizes[:]
genes_copy.append(gene_copy)
genome = Genome(genes_copy, individual.representation,
individual.fitness_func,
individual.natural_fitness)
else:
genome = Genome(individual.genes[:], individual.representation,
individual.fitness_func,
individual.natural_fitness)
genome._fitness = individual._fitness
if hasattr(individual, 'strategy_params') \
and individual.strategy_params is not None:
genome.strategy_params = individual.strategy_params.copy()
if hasattr(individual, 'average_sigmas') \
and individual.average_sigmas is not None:
genome.average_sigmas = individual.average_sigmas[:]
return genome
def pairwise(self, iterable):
"""
Make the given iterable into pairs, i.e.:
s -> (s0,s1), (s2,s3), (s4, s5), ...
"""
a = iter(iterable)
return izip(a, a)
def show_plot(self):
self.plotter.show()
def add_to_plot(self, data, label):
self.plotter.add_to_plot(data, label)