def train(self, train_interval, validate_interval, test_interval):
'''
Args:
train_interval (list[int]) # E.G. (10, 100)
integers - specifying the range (STAR, END) of data we want to use to train the Optimizer
'''
# STEPS:
# 1. Intialize GA parameters
# 2. Set Training/Prediction intervals
# 3. Initialize Evaluator
# 4. Initialize Optimizer
# 5. Train
# pdb.set_trace()
portfolioSize, minWeight, transactionCost, lmbda, predictionInterval, rebalancingPeriod = get_tracker_parameters(
) # we set these paramaters as per the PFGA applciaiton
probCrossover, probReplacement, probMutation, sizeTournament = get_genetics_parameters(
)
self.optimizer = GeneticAlgorithm(
50,
100,
self.stocks,
self.index,
genetic_params=(probCrossover, probReplacement, probMutation,
sizeTournament),
tracker_params=(portfolioSize, minWeight, rebalancingPeriod,
transactionCost, lmbda))
self.optimizer.run(train_interval, validate_interval, test_interval)
def main():
tsp = TSP(5, 50)
ga = GeneticAlgorithm(tsp.population)
for i in range(100000):
ga.generation(0.01)
print ga.best.solution()
def FLNN_GA(X_train,Y_train):
problem_size = 10
pop_size = 10*problem_size
max_gens = 20
search_space = [[-0.5,0.5] for _ in range(problem_size)]
crossf = 0.9
mutationf = 0.005
GA = GeneticAlgorithm(problem_size,pop_size,max_gens,search_space,mutationf,crossf,X_train,Y_train)
best = GA.search()
return best[0]
class TrainOptimizer:
def __init__(self, index, stocks):
'''
Args:
index (Stock)
Index (FTSE-100) represented as a Stock
stocks (List[Stock])
List of stocks containing the assets in the portfolio
'''
# pdb.set_trace()
self.index = index
self.stocks = stocks
def train(self, train_interval, validate_interval, test_interval):
'''
Args:
train_interval (list[int]) # E.G. (10, 100)
integers - specifying the range (STAR, END) of data we want to use to train the Optimizer
'''
# STEPS:
# 1. Intialize GA parameters
# 2. Set Training/Prediction intervals
# 3. Initialize Evaluator
# 4. Initialize Optimizer
# 5. Train
# pdb.set_trace()
portfolioSize, minWeight, transactionCost, lmbda, predictionInterval, rebalancingPeriod = get_tracker_parameters(
) # we set these paramaters as per the PFGA applciaiton
probCrossover, probReplacement, probMutation, sizeTournament = get_genetics_parameters(
)
self.optimizer = GeneticAlgorithm(
50,
100,
self.stocks,
self.index,
genetic_params=(probCrossover, probReplacement, probMutation,
sizeTournament),
tracker_params=(portfolioSize, minWeight, rebalancingPeriod,
transactionCost, lmbda))
self.optimizer.run(train_interval, validate_interval, test_interval)
def best_portfolio(self, datatype='test'):
return self.optimizer.best_portfolio(datatype)
def solution(self):
return self.optimizer.solution()
def crossover_population(self):
crossover_result = []
assert self.population, "Population is empty."
while len(crossover_result) < self.population_size:
route_1 = self.get_route_via_tournament()
route_2 = self.get_route_via_tournament()
crossed_1, crossed_2 = GeneticAlgorithm.crossover(route_1, route_2)
if crossed_1.uniques_only():
crossover_result.append(crossed_1)
if crossed_2.uniques_only():
crossover_result.append(crossed_2)
self.population = crossover_result[:self.population_size]
def main():
parameters = {"lr": [10**-i*j for i, j in product(range(3, 6), [1, 5])],
"dropout": np.arange(0.0, 0.55, .05),
"weight_decay": [10**-i*j
for i, j in product(range(4, 6), [1, 5])],
"optimizer": [func1, func2],
"nb_layers": list(range(2, 21)),
"nb_nodes": [2**i for i in range(1, 11)],
"batch_size": [2**i for i in range(4, 10)]}
# 601920
word = "qwertzu"
# parameters = [list(letters) for _ in range(len(word))]
genetic_algo = GeneticAlgorithm(parameters, 40)
population = genetic_algo.init_population()
print(population)
epochs = 100
final_counter = 0
for _ in range(epochs):
counter = 0
run = True
while run:
accuracies = np.zeros((genetic_algo.population_size, ))
for i, chromosome in enumerate(population):
accuracies[i] = accuracy(word, chromosome)
if accuracies[i] == len(word):
res = "".join(chromosome)
print(f'{res} in {counter} generations')
final_counter += counter
run = False
counter += 1
population = genetic_algo.next_population(accuracies)
print(f'Mean generation: {final_counter/epochs}')
def main(argv):
start = time.time()
parser = argparse.ArgumentParser()
#Creatuer settings
parser.add_argument('--im_size', type=int, default=9, metavar='N',\
help='Size of creature')
parser.add_argument('--voxel_types', type=int, default=5, metavar='N',\
help='How many different types of voxels')
parser.add_argument('--number_neighbors', type=int, default=7, metavar='N',\
help='Number of neighbors')
parser.add_argument('--initial_noise', type=int, default=1, metavar='N',\
help='initial_noise')
# GA setting
parser.add_argument('--sigma', type=float, default = 0.03,\
help='Sigma')
parser.add_argument('--N', type=int, default = 3,\
help='N')
# Voxelyze settings
parser.add_argument('--simtime',type=int,default=0.55,\
help='Simulation time in voxelyze')
parser.add_argument('--initime',type=int,default=0.05,\
help='Intiation time of simulation in voxelyze')
parser.add_argument('--fraction',type=float,default=0.9,\
help='Fraction of the optimal integration step. The lower, the more stable (and slower) the simulation.')
parser.add_argument('--run_directory', type=str, default='basic_data')
parser.add_argument('--run_name', type=str, default='Basic')
#General settings
parser.add_argument('--popsize', type=int, default =10, \
help='Population size.') #3000
parser.add_argument('--generations', type=int, default = 4,\
help='Generations.') #50000
parser.add_argument('--threads', type=int, default=4, metavar='N',\
help='threads')
parser.add_argument('--optimizer', type=str, default='ga', metavar='N',\
help='ga')
parser.add_argument('--recurrent', type=int, default=0, metavar='N',\
help='0 = not recurrent, 1 = recurrent')
parser.add_argument('--data_read', type=int, default=0, metavar='N',\
help='0 = not data read, 1 = dataread')
parser.add_argument('--cell_sleep', type=int, default=0, metavar='N',\
help='0 = not sleep, 0 = sleep')
parser.add_argument('--growth_facter', type=int, default=1, metavar='N',\
help='0 = not growt facter, 1 = growth facter exist')
parser.add_argument('--seed', type=int, default=1, metavar='N',\
help='seed')
parser.add_argument('--folder', type=str, default='results', metavar='N',\
help='folder to store results')
parser.add_argument('--show', type=str, default='none', metavar='N',\
help='visualize genome')
parser.add_argument('--expression', type=int, default=0, metavar='N',\
help='seperate gene expression')
parser.add_argument('--fig_output_rate', type=int, default=1, metavar='N',\
help='fig_output_rate')
args = parser.parse_args()
# You need to add variables that you want to refer to in evaluation.py.
settings = {'voxel_types':args.voxel_types, 'im_size':args.im_size, 'optimizer':args.optimizer, \
'recurrent' : args.recurrent, 'cell_sleep': args.cell_sleep, 'growth_facter':args.growth_facter, 'expression': args.expression, 'seed':args.seed, 'fraction':args.fraction,\
'simtime':args.simtime, 'initime':args.initime, 'run_directory':args.run_directory, 'run_name':args.run_name, 'data_read':args.data_read,\
'number_neighbors':args.number_neighbors, 'sigma':args.sigma, 'N':args.N, 'fig_output_rate':args.fig_output_rate, 'initial_noise':args.initial_noise}
#TODO do not alwaus pass the whole target image
input_dim = settings['number_neighbors']
voxel_types = settings['voxel_types']
if settings['growth_facter']:
input_dim = 3 * 9 * 2
voxel_types = settings['voxel_types'] + 1
if settings['cell_sleep']:
input_dim = settings['number_neighbors'] + 1
voxel_types = settings['voxel_types'] + 1
model = MorphNet(input_dim=input_dim,
number_state=voxel_types,
recurrent=settings['recurrent'])
if args.show != 'none':
x = torch.load(args.show)
ca_fitness(x, True, "iterations.png")
exit()
print("OPTIMIZER ", args.optimizer)
inputfile_path = args.run_directory + "/voxelyzeFiles/"
outputfile_path = args.run_directory + "/fitnessFiles/"
tempfile_path = args.run_directory + "/tempFiles/"
bestocreatures_path = args.run_directory + "/bestofFiles/"
time_path = args.run_directory + "/timeFiles/"
model_path = args.run_directory + "/model_stateFiles/"
dev_states_path = args.run_directory + "/dev_stateFiles/"
try:
if not os.path.exists(inputfile_path):
os.makedirs(inputfile_path)
except OSError as err:
print(err)
try:
if not os.path.exists(outputfile_path):
os.makedirs(outputfile_path)
except OSError as err:
print(err)
try:
if not os.path.exists(tempfile_path):
os.makedirs(tempfile_path)
except OSError as err:
print(err)
try:
if not os.path.exists(bestocreatures_path):
os.makedirs(bestocreatures_path)
except OSError as err:
print(err)
try:
if not os.path.exists(time_path):
os.makedirs(time_path)
except OSError as err:
print(err)
try:
if not os.path.exists(model_path):
os.makedirs(model_path)
except OSError as err:
print(err)
try:
if not os.path.exists(dev_states_path):
os.makedirs(dev_states_path)
except OSError as err:
print(err)
#0.03 worked for evolving heart
if args.optimizer == 'ga':
ga = GeneticAlgorithm(model.get_weights(), run_robot_simulation, population_size=args.popsize, sigma=0.1, \
num_threads=args.threads, folder=args.folder, settings=settings) #0.05 works okay
ga.run(args.generations, print_step=1)
else:
print('No optimizer specified')
process_time = time.time() - start
str_ = str(process_time)
with open("{0}/timeFiles/process_time.txt".format(args.run_directory),
'wt') as f:
f.write(str_)
best_iteration_output += ", "
print(pop_size, "elements")
# repeat multiple time (passage)
time_history = []
iteration_history = []
best_found = 0
best_iteration_history = []
best_time_history = []
for _ in range(passage):
if (alg_type == "ga"):
optimizer_algorithm = GeneticAlgorithm(pop_size,
child_percentage,
tournament_size,
mutation_factor,
plateau_lengh)
if (alg_type == "pso"):
optimizer_algorithm = ParticleSwarmOptimization(
pop_size, intertia_weight, cognitive_weight, social_weight,
plateau_lengh)
start_benchmark = time.time()
best_sample, iteration = optimizer_algorithm.optimize(
lower_bound=lower_bound_list[bench_pos],
upper_bound=upper_bound_list[bench_pos],
function_to_optimize=benchmark_list[bench_pos])
end_benchmark = time.time()
while running and same < 1000:
plotter.ion()
if population is None:
population = FirstPopulation(points)
else:
previous_population = population.get_population()
population = ChildsPopulation(previous_population)
population.crossover_population()
population.mutate_population()
fittest_route = population.get_fittest_route()
if best is None:
best = fittest_route
elif fittest_route.calculate_distance() < best.calculate_distance():
best = fittest_route
plotter.clear()
plotter.draw(points, best.get_full_route())
same = 0
print(Population.number_of_populations, best.calculate_distance())
else:
same += 1
if same > 20:
GeneticAlgorithm.remove_instersection(best)
plotter.show()
plotter.pause()
plotter.draw(points, best.get_full_route())
print(best.get_full_route())
print(best.calculate_distance())
plotter.show()
x4 = x6 = 0
if x6 > x4 and x6 > x5:
x4 = x5 = 0
if x3 == 0:
return 0
else:
return x1 + (10 - (x2 / x3)) + (x4 * 9) + (x5 * 5) + (x6 * 1)
population_size = 100
genome_length = 5
ga = GeneticAlgorithm(fitness_function,
pop_size=population_size,
genome_length=genome_length,
lb=[1, 1, 0, 0, 0],
ub=[20, 20, 2, 2, 2])
ga.generate_binary_population()
# No. of pairs of individuals to be picked to mate
ga.number_of_pairs = 4
# Selective pressure from interval [1.0, 2.0]
# the lower value, the less will the fitness play role
ga.selective_pressure = 1.4
ga.mutation_rate = 0.1
ga.run(100)
import time
import os
import zipfile
POPULATION = 500
GENERATIONS = 10
now = str(datetime.datetime.now())
now = now.replace(' ', '-').replace(':', '').split('.')[0]
os.mkdir(os.getcwd() + os.sep + 'results' + os.sep + now)
t1 = time.time()
grid = Grid((10,10), 10)
opt = GeneticAlgorithm(grid, POPULATION)
print 'Generation %s - Population: %s\n' % (opt.generation, len(opt.population))
t2 = time.time()
print 'Initial population calculation time: %0.3f min' % (float(t2 - t1) / 60.0)
for i in range(1, GENERATIONS + 1):
print '========================================'
print '\nOn generation %s - Population: %s\n' % (opt.generation, len(opt.population))
opt.evolve()
fitness_avg = opt.grade()
avg_repeaters = (sum([len(x[1]) for x in opt.population]) / float(len(opt.population)))
print 'Score of this population: %s' % fitness_avg
def GA_portfolio_optimize(self,
selected_strategy=[
"tesmom", "2nearest", "term_structure",
"hedging_pressure"
],
target="Mean_Variance",
loss_weight=[1, 1, 1]):
# print(selected_strategy)
ret = deepcopy(self.strategy_return_forward)
for test_name_i, backtest_result_i in deepcopy(ret).items():
for method_name_i, return_i in backtest_result_i.items():
if method_name_i not in selected_strategy:
del (ret[test_name_i][method_name_i])
for test_name_i in deepcopy(ret).keys():
if len(ret[test_name_i]) == 0:
del (ret[test_name_i])
self.selected_strategy_return_forward = ret
flag = 0
for test_name_i, backtest_result_i in ret.items():
for method_name_i, return_i in backtest_result_i.items():
return_i = return_i.resample("M").sum()
if flag == 0:
total_ret = pd.DataFrame(
return_i.values,
index=return_i.index,
columns=['{}_{}'.format(test_name_i, method_name_i)])
flag = 1
elif flag == 1:
return_i = pd.DataFrame(
return_i.values,
index=return_i.index,
columns=['{}_{}'.format(test_name_i, method_name_i)])
total_ret = total_ret.join(return_i, how="outer")
# print(total_ret)
total_ret.to_csv('total_ret.csv')
#上面的所有步骤都是为了将数据转化为合适的形式传入到优化函数中
flag = 0
start = time.clock()
counter = 0
total_len = len(total_ret) * 12 - 72
for i in range(1, len(total_ret) + 1):
s = {}
counter += min(i, 12)
ret = total_ret.iloc[max(0, i - 12):i, :]
x = GeneticAlgorithm(ret)
if target == 'test_':
x.run_softmax(target=target, loss_weight=loss_weight)
else:
x.run(target)
_, cleaned_weights = x.best_portfolio("None")
s[total_ret.index[i - 1]] = cleaned_weights
# print(s)
df = pd.DataFrame(s, index=ret.columns)
print(df)
if flag == 0:
final_weights = df
flag = 1
else:
final_weights = final_weights.join(df)
elapsed = (time.clock() - start)
total_time = (elapsed / (counter) * (total_len))
print('Time processed remained : {:.2f}/{:.2f}'.format(
elapsed, total_time))
#得到权重配比的dataframe
final_weights = final_weights.T
self.weights = final_weights
return self.weights
from population import Population
import DQN
import TD3
if RANDOM_SEED:
np.random.seed(RANDOM_SEED)
tf.random.set_random_seed(RANDOM_SEED)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
if 'CartPole' in ENVIRONMENT.name:
if 'dqn' in ALGORITHM:
DQN.train()
elif 'ga' in ALGORITHM:
ga = GeneticAlgorithm()
agents = Population(optimizer=ga)
agents.evolve(save=True)
elif 'es' in ALGORITHM:
es = EvolutionStrategies()
agents = Population(optimizer=es)
agents.evolve(save=True)
else:
print("Please, check that configurations are set correctly.")
elif 'BipedalWalker' in ENVIRONMENT.name:
if 'td3' in ALGORITHM:
agent = TD3.Agent()
agent.train()
elif 'ga' in ALGORITHM:
ga = GeneticAlgorithm()
from GA import GeneticAlgorithm ga = GeneticAlgorithm() ga.run() # pass
##
## city17 = City(200, 40)
## tourManager.add_city(city17)
##
## city18 = City(20, 20)
## tourManager.add_city(city18)
##
## city19 = City(60, 20)
## tourManager.add_city(city19)
##
## city20 = City(160, 20)
## tourManager.add_city(city20)
pop = Population(100)
pop.initia_pop(tourManager)
GA = GeneticAlgorithm(pop, 0.9999, 0.015, 5, 100, True)
#for i in range(100):
# GA.evolution()
#GA.roulette_wheel_selection()
GA.run_test()
#GA.run_test()
#GA.get_best_case()
#GA.check_pop();
def mutate_population(self):
for route in self.population:
GeneticAlgorithm.mutation(route)