def execute(self, front: List[S]) -> List[S]:
if front is None:
raise Exception('The front is null')
elif len(front) == 0:
raise Exception('The front is empty')
ranking = FastNonDominatedRanking(self.dominance_comparator)
ranking.compute_ranking(front)
ranking_index = 0
new_solution_list = []
while len(new_solution_list) < self.max_population_size:
if len(ranking.get_subfront(ranking_index)) < self.max_population_size - len(new_solution_list):
subfront = ranking.get_subfront(ranking_index)
new_solution_list = new_solution_list + subfront
ranking_index += 1
else:
subfront = ranking.get_subfront(ranking_index)
parameter_K = len(subfront) - (self.max_population_size - len(new_solution_list))
while parameter_K > 0:
subfront = self.compute_hypervol_fitness_values(subfront, self.reference_point, parameter_K)
subfront = sorted(subfront, key=lambda x: x.attributes['fitness'], reverse=True)
subfront = subfront[:-1]
parameter_K = parameter_K - 1
new_solution_list = new_solution_list + subfront
return new_solution_list
def execute(self, solution_list: List[S]) -> List[S]:
ranking = FastNonDominatedRanking()
crowding_distance = CrowdingDistance()
ranking.compute_ranking(solution_list)
ranking_index = 0
new_solution_list = []
while len(new_solution_list) < self.max_population_size:
if len(ranking.get_subfront(ranking_index)
) < self.max_population_size - len(new_solution_list):
new_solution_list = new_solution_list + ranking.get_subfront(
ranking_index)
ranking_index += 1
else:
subfront = ranking.get_subfront(ranking_index)
crowding_distance.compute_density_estimator(subfront)
sorted_subfront = sorted(
subfront,
key=lambda x: x.attributes["crowding_distance"],
reverse=True)
for i in range(
(self.max_population_size - len(new_solution_list))):
new_solution_list.append(sorted_subfront[i])
return new_solution_list
def execute(self, front: List[S]) -> List[S]:
if front is None:
raise Exception('The front is null')
elif len(front) == 0:
raise Exception('The front is empty')
ranking = FastNonDominatedRanking(self.dominance_comparator)
crowding_distance = CrowdingDistance()
ranking.compute_ranking(front)
ranking_index = 0
new_solution_list = []
while len(new_solution_list) < self.max_population_size:
if len(ranking.get_subfront(ranking_index)) < (self.max_population_size - len(new_solution_list)):
new_solution_list = new_solution_list + ranking.get_subfront(ranking_index)
ranking_index += 1
else:
subfront = ranking.get_subfront(ranking_index)
crowding_distance.compute_density_estimator(subfront)
sorted_subfront = sorted(subfront, key=lambda x: x.attributes['crowding_distance'], reverse=True)
for i in range((self.max_population_size - len(new_solution_list))):
new_solution_list.append(sorted_subfront[i])
return new_solution_list
def replacement(self, population: List[S],
offspring_population: List[S]) -> List[S]:
""" Implements NSGA-III selection as described in
* Deb, K., & Jain, H. (2014). An Evolutionary Many-Objective Optimization
Algorithm Using Reference-Point-Based Nondominated Sorting Approach,
Part I: Solving Problems With Box Constraints. IEEE Transactions on
Evolutionary Computation, 18(4), 577–601. doi:10.1109/TEVC.2013.2281535.
"""
# Algorithm 1 steps 4--8
ranking = FastNonDominatedRanking(self.dominance_comparator)
ranking.compute_ranking(population + offspring_population)
ranking_index = 0
pop = []
while len(pop) < self.population_size:
if len(ranking.get_subfront(
ranking_index)) < self.population_size - len(pop):
pop += ranking.get_subfront(ranking_index)
ranking_index += 1
else:
break
# complete selected individuals using the reference point based approach
selection = EnvironmentalSelection(
number_of_objectives=self.problem.number_of_objectives,
k=self.population_size - len(pop))
pop += selection.execute(ranking.get_subfront(ranking_index))
return pop
def update_external_archive(self):
feasible_solutions = []
for solution in self.solutions:
if is_feasible(solution):
feasible_solutions.append(copy.deepcopy(solution))
if len(feasible_solutions) > 0:
feasible_solutions = feasible_solutions + self.archive
ranking = FastNonDominatedRanking()
ranking.compute_ranking(feasible_solutions)
first_rank_solutions = ranking.get_subfront(0)
if len(first_rank_solutions) <= self.population_size:
self.archive = []
for solution in first_rank_solutions:
self.archive.append(copy.deepcopy(solution))
else:
crowding_distance = CrowdingDistance()
while len(first_rank_solutions) > self.population_size:
crowding_distance.compute_density_estimator(
first_rank_solutions)
first_rank_solutions = sorted(
first_rank_solutions,
key=lambda x: x.attributes['crowding_distance'],
reverse=True)
first_rank_solutions.pop()
self.archive = []
for solution in first_rank_solutions:
self.archive.append(copy.deepcopy(solution))
def get_result(self):
""" Return only non dominated solutions."""
ranking = FastNonDominatedRanking(self.dominance_comparator)
ranking.compute_ranking(self.solutions, k=self.population_size)
for solution in self.solutions:
# print(solution.objectives, self.target_pattern)
goal_flag = np.zeros((7), dtype=int)
for j in range(7):
if solution.objectives[j] < self.target_value_threshold[j]:
goal_flag[j] = 1
else:
goal_flag[j] = 0
if (goal_flag == np.array(self.target_pattern)).all():
# print(goal_flag, self.target_pattern)
self.problem_solved = True
return ranking.get_subfront(0)
def replacement(self, population: List[S], offspring_population: List[S]) -> List[S]:
# print("replacement")
""" Implements NSGA-III environmental selection based on reference points as described in:
* Deb, K., & Jain, H. (2014). An Evolutionary Many-Objective Optimization
Algorithm Using Reference-Point-Based Nondominated Sorting Approach,
Part I: Solving Problems With Box Constraints. IEEE Transactions on
Evolutionary Computation, 18(4), 577–601. doi:10.1109/TEVC.2013.2281535.
"""
F = np.array([s.objectives for s in population])
# find or usually update the new ideal point - from feasible solutions
# note that we are assuming minimization here!
self.ideal_point = np.min(np.vstack((self.ideal_point, F)), axis=0)
self.worst_point = np.max(np.vstack((self.worst_point, F)), axis=0)
# calculate the fronts of the population
ranking = FastNonDominatedRanking(self.dominance_comparator)
ranking.compute_ranking(population + offspring_population, k=self.population_size)
fronts, non_dominated = ranking.ranked_sublists, ranking.get_subfront(0)
# find the extreme points for normalization
self.extreme_points = get_extreme_points(F=np.array([s.objectives for s in non_dominated]),
n_objs=self.problem.number_of_objectives,
ideal_point=self.ideal_point,
extreme_points=self.extreme_points)
# find the intercepts for normalization and do backup if gaussian elimination fails
worst_of_population = np.max(F, axis=0)
worst_of_front = np.max(np.array([s.objectives for s in non_dominated]), axis=0)
nadir_point = get_nadir_point(extreme_points=self.extreme_points,
ideal_point=self.ideal_point,
worst_point=self.worst_point,
worst_of_population=worst_of_population,
worst_of_front=worst_of_front)
# consider only the population until we come to the splitting front
pop = np.concatenate(ranking.ranked_sublists)
F = np.array([s.objectives for s in pop])
# update the front indices for the current population
counter = 0
for i in range(len(fronts)):
for j in range(len(fronts[i])):
fronts[i][j] = counter
counter += 1
last_front = np.array(fronts[-1])
# associate individuals to niches
niche_of_individuals, dist_to_niche = associate_to_niches(F=F,
niches=self.reference_directions,
ideal_point=self.ideal_point,
nadir_point=nadir_point)
# if we need to select individuals to survive
if len(pop) > self.population_size:
# if there is only one front
if len(fronts) == 1:
until_last_front = np.array([], dtype=np.int)
niche_count = np.zeros(len(self.reference_directions), dtype=np.int)
n_remaining = self.population_size
# if some individuals already survived
else:
until_last_front = np.concatenate(fronts[:-1])
niche_count = compute_niche_count(len(self.reference_directions),
niche_of_individuals[until_last_front])
n_remaining = self.population_size - len(until_last_front)
S_idx = niching(pop=pop[last_front],
n_remaining=n_remaining,
niche_count=niche_count,
niche_of_individuals=niche_of_individuals[last_front],
dist_to_niche=dist_to_niche[last_front])
survivors_idx = np.concatenate((until_last_front, last_front[S_idx].tolist()))
pop = pop[survivors_idx]
return list(pop)
def get_result(self):
""" Return only non dominated solutions."""
ranking = FastNonDominatedRanking(self.dominance_comparator)
ranking.compute_ranking(self.solutions, k=self.population_size)
return ranking.get_subfront(0)
class FastNonDominatedRankingTestCases(unittest.TestCase):
def setUp(self):
self.ranking = FastNonDominatedRanking()
def test_should_constructor_create_a_valid_object(self):
self.assertIsNotNone(self.ranking)
def test_should_compute_ranking_of_an_emtpy_solution_list_return_a_empty_list_of_subranks(self):
solution_list = []
self.assertEqual(0, len(self.ranking.compute_ranking(solution_list)))
def test_should_compute_ranking_return_a_subfront_if_the_solution_list_contains_one_solution(self):
solution = Solution(2, 3)
solution_list = [solution]
ranking = self.ranking.compute_ranking(solution_list)
self.assertEqual(1, self.ranking.get_number_of_subfronts())
self.assertEqual(solution, ranking[0][0])
def test_should_compute_ranking_return_a_subfront_if_the_solution_list_contains_two_nondominated_solutions(self):
solution = Solution(2, 2)
solution.objectives = [1, 2]
solution2 = Solution(2, 2)
solution2.objectives = [2, 1]
solution_list = [solution, solution2]
ranking = self.ranking.compute_ranking(solution_list)
self.assertEqual(1, self.ranking.get_number_of_subfronts())
self.assertEqual(2, len(self.ranking.get_subfront(0)))
self.assertEqual(solution, ranking[0][0])
self.assertEqual(solution2, ranking[0][1])
def test_should_compute_ranking_work_properly_case1(self):
""" The list contains two solutions and one of them is dominated by the other one.
"""
solution = Solution(2, 2)
solution.objectives = [2, 3]
solution2 = Solution(2, 2)
solution2.objectives = [3, 6]
solution_list = [solution, solution2]
ranking = self.ranking.compute_ranking(solution_list)
self.assertEqual(2, self.ranking.get_number_of_subfronts())
self.assertEqual(1, len(self.ranking.get_subfront(0)))
self.assertEqual(1, len(self.ranking.get_subfront(1)))
self.assertEqual(solution, ranking[0][0])
self.assertEqual(solution2, ranking[1][0])
def test_should_ranking_of_a_population_with_three_dominated_solutions_return_three_subfronts(self):
solution = Solution(2, 2)
solution.objectives = [2, 3]
solution2 = Solution(2, 2)
solution2.objectives = [3, 6]
solution3 = Solution(2, 2)
solution3.objectives = [4, 8]
solution_list = [solution, solution2, solution3]
ranking = self.ranking.compute_ranking(solution_list)
self.assertEqual(3, self.ranking.get_number_of_subfronts())
self.assertEqual(1, len(self.ranking.get_subfront(0)))
self.assertEqual(1, len(self.ranking.get_subfront(1)))
self.assertEqual(1, len(self.ranking.get_subfront(2)))
self.assertEqual(solution, ranking[0][0])
self.assertEqual(solution2, ranking[1][0])
self.assertEqual(solution3, ranking[2][0])
def test_should_ranking_of_a_population_with_five_solutions_work_properly(self):
solution1 = Solution(2, 2)
solution2 = Solution(2, 2)
solution3 = Solution(2, 2)
solution4 = Solution(2, 2)
solution5 = Solution(2, 2)
solution1.objectives[0] = 1.0
solution1.objectives[1] = 0.0
solution2.objectives[0] = 0.5
solution2.objectives[1] = 0.5
solution3.objectives[0] = 0.0
solution3.objectives[1] = 1.0
solution4.objectives[0] = 0.6
solution4.objectives[1] = 0.6
solution5.objectives[0] = 0.7
solution5.objectives[1] = 0.5
solutions = [solution1, solution2, solution3, solution4, solution5]
ranking = self.ranking.compute_ranking(solutions)
self.assertEqual(2, self.ranking.get_number_of_subfronts())
self.assertEqual(3, len(self.ranking.get_subfront(0)))
self.assertEqual(2, len(self.ranking.get_subfront(1)))
self.assertEqual(solution1, ranking[0][0])
self.assertEqual(solution2, ranking[0][1])
self.assertEqual(solution3, ranking[0][2])
self.assertEqual(solution4, ranking[1][0])
self.assertEqual(solution5, ranking[1][1])
def dtlz_test(p: FloatProblemGD, label: str = '', experiment: int = 50):
# problem.reference_front = read_solutions(filename='resources/reference_front/DTLZ2.3D.pf')
max_evaluations = 25000
# references = ReferenceDirectionFromSolution(p)
algorithm = NSGA3C(
problem=p,
population_size=100,
reference_directions=UniformReferenceDirectionFactory(p.instance_.n_obj, n_points=92),
mutation=PolynomialMutation(probability=1.0 / p.number_of_variables, distribution_index=20),
crossover=SBXCrossover(probability=1.0, distribution_index=30),
termination_criterion=StoppingByEvaluations(max_evaluations=max_evaluations)
)
bag = []
total_time = 0
for i in range(experiment):
algorithm = NSGA3C(
problem=p,
population_size=92,
reference_directions=UniformReferenceDirectionFactory(p.instance_.n_obj, n_points=91),
mutation=PolynomialMutation(probability=1.0 / p.number_of_variables, distribution_index=20),
crossover=SBXCrossover(probability=1.0, distribution_index=30),
termination_criterion=StoppingByEvaluations(max_evaluations=max_evaluations)
)
progress_bar = ProgressBarObserver(max=max_evaluations)
algorithm.observable.register(progress_bar)
algorithm.run()
total_time += algorithm.total_computing_time
bag = bag + algorithm.get_result()
print(len(bag))
print('Total computing time:', total_time)
print('Average time: ', str(total_time / experiment))
print_solutions_to_file(bag, DIRECTORY_RESULTS + 'Solutions.bag._class_' + label + algorithm.label)
ranking = FastNonDominatedRanking()
ranking.compute_ranking(bag)
front_ = ranking.get_subfront(0)
print('Front 0 size : ', len(front_))
alabels = []
for obj in range(p.number_of_objectives):
alabels.append('Obj-' + str(obj))
plot_front = Plot(title='Pareto front approximation' + ' ' + label,
axis_labels=alabels)
plot_front.plot(front_, label=label + 'F0 ' + algorithm.label,
filename=DIRECTORY_RESULTS + 'F0_class_' + 'original_' + label + algorithm.label,
format='png')
class_fronts = [[], [], [], []]
for s in front_:
_class = problem.classifier.classify(s)
if _class[0] > 0:
class_fronts[0].append(s)
elif _class[1] > 0:
class_fronts[1].append(s)
elif _class[2] > 0:
class_fronts[2].append(s)
else:
class_fronts[3].append(s)
print(len(class_fronts[0]), len(class_fronts[1]), len(class_fronts[2]), len(class_fronts[3]))
_front = class_fronts[0] + class_fronts[1]
if len(_front) == 0:
_front = class_fronts[2] + class_fronts[3]
print('Class : ', len(_front))
# Save results to file
print_solutions_to_file(_front, DIRECTORY_RESULTS + 'Class_F0' + label + algorithm.label)
print(f'Algorithm: ${algorithm.get_name()}')
print(f'Problem: ${p.get_name()}')
plot_front = Plot(title=label + 'F_' + p.get_name(), axis_labels=alabels)
plot_front.plot(_front, label=label + 'F_' + p.get_name(),
filename=DIRECTORY_RESULTS + 'Class_F0' + label + p.get_name(),
format='png')
def validate_interclass(problem):
with open('/home/thinkpad/Documents/jemoa/experiments/dtlz_preferences/Class_F0DTLZ1_P3.out') as f:
content = f.readlines()
# you may also want to remove whitespace characters like `\n` at the end of each line
content = [x.strip() for x in content]
solutions = []
print(problem.get_preference_model(0))
for idx, line in enumerate(content):
split = line.split('*')[1].split(',')
_s = problem.generate_existing_solution([float(x) for x in split],is_objectives=True)
_s.bag = 'java'
solutions.append(_s)
with open('/home/thinkpad/PycharmProjects/jMetalPy/results/Class_F0enfoque_frontsNSGAIII_custom.DTLZ1P_3.csv') as f:
content = f.readlines()
# you may also want to remove whitespace characters like `\n` at the end of each line
content = [x.strip() for x in content]
for idx, line in enumerate(content):
if not 'variables' in line:
split = line.split('*')[1].split(',')
_s = problem.generate_existing_solution([float(x) for x in split], is_objectives=True)
_s.bag = 'python'
solutions.append(_s)
ranking = FastNonDominatedRanking()
ranking.compute_ranking(solutions)
front_ = ranking.get_subfront(0)
print('Solutions:',len(solutions))
print('Front 0:', len(front_))
class_fronts = [[], [], [], []]
fjava = 0
fpython = 0
classifier = InterClassNC(problem)
for s in front_:
if s.bag == 'java':
fjava += 1
else:
fpython += 1
# problem.evaluate(s)
_class = classifier.classify(s)
if _class[0] > 0:
class_fronts[0].append(s)
elif _class[1] > 0:
class_fronts[1].append(s)
elif _class[2] > 0:
class_fronts[2].append(s)
else:
class_fronts[3].append(s)
print('Java solutions:', (fjava / len(front_)))
print('Python solutions:', (fpython / len(front_)))
print('HSat : ', len(class_fronts[0]), ', Sat : ', len(class_fronts[1]), ', Dis : ', len(class_fronts[2]),
', HDis : ', len(class_fronts[3]))
_sat = class_fronts[0] + class_fronts[1]
fjava = 0
fpython = 0
for s in _sat:
if s.bag == 'java':
fjava += 1
else:
fpython += 1
print('Sat Java solutions:', (fjava / len(_sat)))
print('Sat Python solutions:', (fpython / len(_sat)))
plot_front = Plot(title='Sat and HSat Front')
plot_front.plot(_sat, label= 'Sat and HSat Front',
filename=DIRECTORY_RESULTS + 'SatFront' + problem.get_name(),
format='png')
problem.number_of_variables,
distribution_index=20),
crossover=SBXCrossover(probability=1.0, distribution_index=30),
termination_criterion=StoppingByEvaluations(
max_evaluations=max_evaluations))
progress_bar = ProgressBarObserver(max=max_evaluations)
algorithm.observable.register(progress_bar)
algorithm.run()
bag += algorithm.get_result()
print(len(bag))
print_solutions_to_file(
bag, DIRECTORY_RESULTS + 'Solutions.bag.' + algorithm.label)
ranking = FastNonDominatedRanking()
ranking.compute_ranking(bag)
front = ranking.get_subfront(0)
print(len(front))
# Save results to file
print_solutions_to_file(front,
DIRECTORY_RESULTS + 'front0.' + algorithm.label)
plot_front = Plot(title='Pareto front approximation',
axis_labels=['x', 'y', 'z'])
plot_front.plot(front,
label='NSGAII-ZDT7',
filename=DIRECTORY_RESULTS + 'NSGAII-ZDT1P',
format='png')
print(f'Algorithm: ${algorithm.get_name()}')
print(f'Problem: ${problem.get_name()}')
print(f'Computing time: ${algorithm.total_computing_time}')
def get_result(self):
ranking = FastNonDominatedRanking(self.dominance_comparator)
ranking.compute_ranking(self.solutions)
return ranking.get_subfront(0)
