def __init__(self, objfunc, var_bounds, individual_size, max_iter,
max_or_min, **kwargs):
super().__init__(objfunc)
self.max_iter = max_iter
# 定义个体 / 种群
self.individual = BinaryIndividual(ranges=var_bounds, eps=0.001)
self.population = Population(indv_template=self.individual,
size=individual_size).init()
# Create genetic operators.
# selection = RouletteWheelSelection()
selection = TournamentSelection()
crossover = UniformCrossover(pc=0.8, pe=0.5)
mutation = FlipBitBigMutation(pm=0.1, pbm=0.55, alpha=0.6)
self.engine = GAEngine(population=self.population,
selection=selection,
crossover=crossover,
mutation=mutation,
analysis=[FitnessStore])
@self.engine.fitness_register
def fitness(indv):
"""
适应度函数： 注意这里默认为优化得到最小值
:param indv:
:return:
"""
x = indv.solution
if max_or_min == 'max':
return objfunc(x, **kwargs)
else:
return -objfunc(x, **kwargs)
@self.engine.analysis_register
class ConsoleOutputAnalysis(OnTheFlyAnalysis):
interval = 1
master_only = True
def register_step(self, g, population, engine):
best_indv = population.best_indv(engine.fitness)
msg = 'Generation: {}, best fitness: {:.3f}'.format(
g, engine.fitness(best_indv))
# self.logger.info(msg)
def finalize(self, population, engine):
best_indv = population.best_indv(engine.fitness)
x = best_indv.solution
y = engine.fitness(best_indv)
msg = 'Optimal solution: ({}, {})'.format(x, y)
def tune_weights(self):
old_fitness = self.individual.fitness
weights_scaling = self.individual.get_subtree_scaling()
weights_translation = self.individual.get_subtree_translation()
# Create array with range for each scaling and translation parameter
range = [
self.scale_range,
] * len(weights_scaling) + [
self.translation_range,
] * len(weights_translation)
indv_template = DecimalIndividual(ranges=range, eps=0.1)
population = Population(indv_template=indv_template,
size=self.pop_size)
population.init()
engine = GAEngine(
population=population,
selection=TournamentSelection(),
crossover=GaussianCrossover(pc=1.0),
mutation=NoMutation(),
fitness=self.fitness_function_GAFT,
analysis=[
new_early_stopping_analysis(scale_range=self.scale_range)
])
engine.logger = NoLoggingLogger()
# Run the GA with the specified number of iterations
try:
engine.run(ng=self.max_iterations)
except ValueError:
pass
# Get the best individual.
best_indv = engine.population.best_indv(engine.fitness)
# Log the tuning process
print(
f"Tuner {np.round(old_fitness, 3)} {np.round(-engine.ori_fmax, 3)}"
)
# Only use the new individual if it was really improved
if old_fitness > -engine.ori_fmax:
weights_scaling, weights_translation = self.split_list(
best_indv.solution)
self.individual.set_subtree_scaling(weights_scaling)
self.individual.set_subtree_translation(weights_translation)
self.individual.fitness = -engine.ori_fmax
return deepcopy(self.individual)
def __init__(self ,k ,total_implied_variance ,slice_before ,slice_after ,tau):
self.k =k
self.total_implied_variance =total_implied_variance
self.slice_before =slice_before
self.slice_after = slice_after
self.tau = tau
# Define population.
indv_template = BinaryIndividual(ranges=[(1e-5, 20),(1e-5, 20),(1e-5, 20)], eps=0.001)
self.population = Population(indv_template=indv_template, size=30).init()
# Create genetic operators.
selection = TournamentSelection()
crossover = UniformCrossover(pc=0.8, pe=0.5)
mutation = FlipBitMutation(pm=0.1)
# Create genetic algorithm engine.
self.engine = GAEngine(population=self.population, selection=selection,
crossover=crossover, mutation=mutation,
analysis=[FitnessStore])
# Define fitness function.
@self.engine.fitness_register
@self.engine.minimize
def fitness(indv):
a, b, m, rho, sigma = indv.solution
model_total_implied_variance=svi_raw(self.k,np.array([a, b, m, rho, sigma]),self.tau)
value = norm(self.total_implied_variance - model_total_implied_variance,ord=2)
# if bool(len(self.slice_before)) and np.array(model_total_implied_variance < self.slice_before).any():
# value +=(np.count_nonzero(~np.array(model_total_implied_variance < self.slice_before))*100)
# # value = 1e6
#
# if bool(len(self.slice_after)) and np.array(model_total_implied_variance > self.slice_after).any():
# value += float(np.count_nonzero(~np.array(model_total_implied_variance > self.slice_after)) * 100)
# # value = 1e6
# if np.isnan(value):
# value = 1e6
value = float(value)
return value
from gaft.analysis.fitness_store import FitnessStore
# Define population.
# 先对你所指定的初始种群进行编码
indv_template = BinaryIndividual(ranges=[(0, 10)], eps=0.001)
'''
:param ranges: value ranges for all entries in solution.
:type ranges: list of range tuples. e.g. [(0, 1), (-1, 1)]
:param eps: decrete precisions for binary encoding, default is 0.001.
:type eps: float or float list with the same length with ranges.
'''
population = Population(indv_template=indv_template, size=30).init()
# Create genetic operators.
selection = TournamentSelection()
crossover = UniformCrossover(pc=0.8, pe=0.5)
'''
Crossover operator with uniform crossover algorithm,
see https://en.wikipedia.org/wiki/Crossover_(genetic_algorithm)
:param pc: The probability of crossover (usaully between 0.25 ~ 1.0)
:type pc: float in (0.0, 1.0]
:param pe: Gene exchange probability.
'''
mutation = FlipBitMutation(pm=0.1)
# pm is the possibility of the mutation
# Create genetic algorithm engine.
engine = GAEngine(population=population, selection=selection,
crossover=crossover, mutation=mutation,
fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(geracoes, n_nines)
plt.show()
if __name__ == "__main__":
individuo = BinaryIndividual(ranges=[(-100, 100), (-100, 100)],
eps=0.000001)
populacao = Population(indv_template=individuo, size=100)
populacao.init()
selecao = TournamentSelection()
crossover = UniformCrossover(pc=0.65, pe=0.65)
mutacao = FlipBitMutation(pm=0.008)
engine = GAEngine(population=populacao,
selection=selecao,
crossover=crossover,
mutation=mutacao,
analysis=[FitnessStore, ConsoleOutput])
@engine.fitness_register
def aptidao(ind):
x, y = ind.solution
return 0.5 - ((sin(sqrt(x**2 + y**2))**2 - 0.5) / (1 + 0.001 *
(x**2 + y**2))**2)
def generate():
import random
good_seed = int(sys.argv[1])
print('currrrrrrrrrrrrrrrrrrrrrrrrrrrrr', good_seed)
envSeqDec = ChallengeProveEnvironment() # Initialise a New Challenge Environment to post entire policy
#env = ChallengeEnvironment(experimentCount = 20000)
eps = 0.1 # equal to actions space resolution # range/eps
pop_size = 4
cross_prob = 0.6
exchange_prob = 0.7
mutation_pob = 0.8
generation = 4
REWARDS = []
NEW = []
POLICY = []
tmp_reward = []
tmp_policy = []
policy_450 = []
reward_generation = []
time = []
random.seed(good_seed)
# best_action = ([0], [0.8, 1])
turb = 0
test_action = ([[i/10 for i in range(0, 11)],[i/10 for i in range(0,11)]])
# test_action = ([0, 0.1, 0.2, 0.3, 0.4, 0.5], [0.6, 0.7, 0.8, 0.9, 1])
# 2. Define population
indv_template = OrderIndividual(ranges=[(0, 1), (0, 1), (0, 1), (0, 1), (0, 1), (0, 1), (0, 1), (0, 1), (0, 1), (0, 1)], eps=eps, actions = test_action) # low_bit and high_bit
population = Population(indv_template=indv_template, size = pop_size)
population.init() # Initialize population with individuals.
# 3. Create genetic operators
# Use built-in operators here.
#selection = RouletteWheelSelection()
selection = TournamentSelection()
crossover = UniformCrossover(pc=cross_prob, pe=exchange_prob) # PE = Gene exchange probability
mutation = FlipBitMutation(pm=mutation_pob) # 0.1 todo The probability of mutation
# 4. Create genetic algorithm engine to run optimization
engine = GAEngine(population=population, selection=selection,
crossover=crossover, mutation=mutation,)
# analysis=[FitnessStore])
# 5. Define and register fitness function
@engine.fitness_register
#@engine.dynamic_linear_scaling(target='max', ksi0=2, r=0.9)
def fitness(indv):
p = [0 for _ in range(10)]
p = indv.solution
# encode
policy = {'1': [p[0], p[1]], '2': [p[2], p[3]], '3': [p[4], p[5]], '4': [p[6], p[7]], '5': [p[8], p[9]]}
reward = envSeqDec.evaluatePolicy(policy) # Action in Year 1 only
#print('Sequential Result : ', reward)
tmp_reward.append(reward)
reward_generation.append(reward)
tmp_policy.append(policy)
#print('Policy : ', policy)
#print(policy_450,'**************************good solution***************')
#print(policy_bad,'**************************bad solution***************')
return reward + uniform(-turb, turb)
@engine.analysis_register
class ConsoleOutput(OnTheFlyAnalysis):
master_only = True
interval = 1
def register_step(self, g, population, engine):
best_indv = population.best_indv(engine.fitness)
msg = 'Generation: {}, best fitness: {:.3f}'.format(g + 1, engine.fmax)
#best_reward = max(tmp_reward[g + pop_size * (generation - 1): g + pop_size * generation])
#print(pop_size * (g - 0), pop_size * (g + 1),'&&&&&&&&&&&&&&&&&&&&&&&&&&&&&')
REWARDS.append(max(reward_generation[pop_size * (g - 0): pop_size * (g + 1)]))
#best_policy = POLICY[tmp_reward.index(best_reward)]
#POLICY.append(best_policy)
engine.logger.info(msg)
engine.run(ng = generation)
# print(policy_450)
x = list(range(len(REWARDS)))
plt.plot(x, REWARDS)
plt.title(f'Sequential Rewards {good_seed}')
plt.savefig(f'./GA_trick/GA_seed_{good_seed}.jpg')
# #plt.savefig(f'./res_geneticAlgorithm/Sequential_Rewards_eps:{eps}_popsize:{pop_size}_generation:{generation}_mutation_pob:{mutation_pob}_exchange_prob:{exchange_prob}_cross_prob:{cross_prob}.jpg')
plt.show()
obsSOM, true_labels = utils.loadSOM(
save_dir=config.OUTPUT_TRAINED_MODELS_PATH,
file_name=config.ZSEL_SOM_3D_MODEL_NAME)
setupEnv()
# print(MODELS_VALUES.shape)
# print(MODELS_DICT)
indv_template = DecimalIndividual(ranges=[
(0, 1) for _ in range(config.NB_MODELS)
],
eps=0.001)
population = Population(indv_template=indv_template, size=POPULATION_SIZE)
population.init()
selection = TournamentSelection(
tournament_size=10) # RouletteWheelSelection()
crossover = UniformCrossover(pc=CROSSOVER_PROBABILITY, pe=GE_PROBABILITY)
mutation = FlipBitMutation(pm=MUTATION_PROBABILITY)
engine = GAEngine(population=population,
selection=selection,
crossover=crossover,
mutation=mutation,
analysis=[FitnessStore])
@engine.fitness_register
def fitness(indv):
global MODELS_VALUES
global MODELS_DICT
global CASE
global true_labels
