def _placement_algorithm_execute(self, scenario, reqid):
"""Executes the placement algorithm for the given scenario.If for some reason
the algorithm fails to produce a valid placement (e.g., some constraint is violated),
the function still returns the placement, but specific error information is included
inside the JSON object which represents the solution.
"""
cfg = copy.deepcopy(self.configuration)
cfg["scenario"] = scenario
g = GA(cfg)
if g.error:
err = g.error_string
solution = copy.deepcopy(scenario)
else:
solution = g.execute()
# Check for errors/constrain violations
err = None
if not solution["solution_performance"]["link_capacity_constraints_ok"]:
err = "Link capacity exceeded."
if not solution["solution_performance"]["delay_constraints_ok"]:
err = "Delay constraints violated."
if not solution["solution_performance"]["host_capacity_constraints_ok"]:
err = "Host capacity constraints violated."
if not solution["solution_performance"]["legal_placement"]:
err = "Illegal placement."
self._output_solution(solution)
# update the record in the database with the new status and the derived solution
timestamp = datetime.datetime.now().strftime("%Y-%m-%dT%H:%M:%S.%f")
if err is not None:
status = "FAIL"
else:
status = "SUCCESS"
req = self.db_collection.find_one_and_update(
{"ReqId": reqid},
{'$set': {"finished": timestamp, "solution": solution, "status": status, "info": err}},
projection={"_id": False},
return_document=pymongo.collection.ReturnDocument.AFTER)
# translate solution to the external format
translated_solution = translator.translate_solution(solution)
req_ext = self.db_collection_external.find_one_and_update(
{"ReqId": reqid},
{'$set': {"finished": timestamp, "solution": translated_solution, "status": status, "info": err}},
projection={"_id": False},
return_document=pymongo.collection.ReturnDocument.AFTER)
return translated_solution, err
class GaTestCase(TestCase):
def setUp(self):
self.fake_iter_plugin = FakeIterPlugin()
self.fake_codec_plugin = FakeCodecPlugin()
self.ga = GA(
[self.fake_codec_plugin],
[], [self.fake_iter_plugin], "")
def testSetup(self):
fake_population = FakePopulation()
fake_selector = FakeSelector()
self.ga.setup_population(fake_population)
self.ga.use_selector(fake_selector)
self.assertEqual(self.ga.population, fake_population)
self.assertEqual(fake_population.selector, fake_selector)
# run next
self.ga.next()
self.assertEqual(self.fake_iter_plugin.ga, self.ga)
self.assertEqual(
fake_population.codec_plugin, [self.fake_codec_plugin])
def setUp(self):
self.fake_iter_plugin = FakeIterPlugin()
self.fake_codec_plugin = FakeCodecPlugin()
self.ga = GA(
[self.fake_codec_plugin],
[], [self.fake_iter_plugin], "")
net, train_loss, test_loss = trainBP(x_train_scaled.shape[1], hidden_number, 1,
learning_rate, epochs, shuffle_seed, BATCH_SIZE, loss_fn,
x_train_scaled, y_train_scaled,
x_test_scaled, y_test_scaled)
plotBP(train_loss, test_loss)
plot_net_predict_result('BP ', net, x_test_scaled, target_sc, y_test)
#遗传算法参数
popsize = 100 #遗传算法种群数
iter_max = 6 #遗传算法迭代次数
PM = 0.05 #变异概率
PC = 0.7 #交叉概率
input_number = 21#输入维度
output_number = 1#输出维度
# 自适应参数
PM1 = 0.05 # 变异概率下限
PM2 = 0.25 # 变异概率上限
PC1 = 0.5 # 交叉概率下限
PC2 = 0.8 # 交叉概率上限
adga = AGABP(net, input_number, hidden_number, output_number, popsize, iter_max, PM, PC, None,
PM1, PM2, PC1, PC2)
bestChrom_ga, bestValue_ga, tracematga = adga.adgafun(x_test_scaled, y_test, target_sc)
aga_net = adga.recover_net(bestChrom_ga)
plot_net_predict_result("AGA ", aga_net, x_test_scaled, target_sc, y_test)
gaObject = GA(net, input_number, hidden_number, output_number, popsize, iter_max, PM, PC)
bestChrom_ga, bestValue_ga, tracematga = gaObject.gafun(x_train_scaled, y_train_scaled, target_sc)
ga_net = gaObject.recover_net(bestChrom_ga)
plot_net_predict_result("GA ", ga_net, x_test_scaled, target_sc, y_test)