def runGA(vector):
algorithm_param = {'max_num_iteration': 100,\
'population_size': 500,\
'mutation_probability': .6,\
'elit_ratio': .02,\
'crossover_probability': .1,\
'parents_portion': .4,\
'crossover_type':'uniform',\
'max_iteration_without_improv':50}
varbound =np.array([[0,1],[1,4],[0,0.8],[0,1],
[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]])
#[V_gBurn,ng,debt_level,V_cng_p,farm1,farm2,farm3,farm4,farm5,farm6,farm7]
start = timeit.default_timer()
var_type = np.array([['real'],['int'],['real'],['real'],['real'],
['int'],['int'],['int'],['int'],['int'],['int'],['int']])
model2=ga(function=biodigestor,\
dimension=len(vector),\
variable_type_mixed=var_type,\
variable_boundaries=varbound,\
function_timeout =600,\
algorithm_parameters=algorithm_param)
model2.run()
stop = timeit.default_timer()
print('Run time: '+str(stop-start)+' second')
return model2
def genetic(encodings, category):
cnn = DenseNet169(include_top=False, input_shape=(224, 224, 3))
image_classifier = load_model('image_model.h5')
def f(X):
sum = X[0] * encodings[0]
for i in range(1, len(X)):
sum += X[i] * encodings[i]
decoded_image = (decode_image(sum) * 255).astype(np.uint8)
decoded_image = Image.fromarray(decoded_image)
decoded_image = decoded_image.resize((224, 224))
decoded_image = np.expand_dims(decoded_image, axis=0)
decoded_image = preprocess_input(decoded_image)
features = cnn.predict(decoded_image)
answer = image_classifier.predict(features)[0][category]
return -answer
varbound = np.array([[0, 1]] * 10)
model = ga(function=f,
dimension=10,
variable_type='real',
variable_boundaries=varbound)
model.run()
return model.output_dict['variable']
def genetic_tcn():
"""TCN input/architecture genetic optimization"""
varbound = np.array([
[1, 6], #histo lag 3, 4, 8, 16, 32, 64
[4, 6], #nb_filters 16, 32, 64
[2, 8], #kernel size from 2 to 8
[1, 3] #nb_stacks
])
algorithm_parameters = {
'max_num_iteration': None,
'population_size': 100, #default is 100
'mutation_probability': 0.1,
'elit_ratio': 0.01,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'uniform',
'max_iteration_without_improv': None
}
optim = ga(function=rmse_tcn,
dimension=4,
variable_type='int',
variable_boundaries=varbound,
function_timeout=60 * 60,
algorithm_parameters=algorithm_parameters)
optim.run()
def calc_genetic(num, iterations, d):
filename = "lab2_files/" + str(num) + ".txt"
print(filename)
n, m, data, abs_result = read_data_from_file_2(filename)
def f(x):
predict_y = list(map(lambda row: predict_with_one2(row, x), data))
smape = nrmse(abs_result, predict_y)
return smape
ln = m + 1
varbound = np.array([[-d, d]] * ln)
# algorithm_param = {'max_num_iteration': None,
algorithm_param = {'max_num_iteration': iterations,
'population_size': 100,
'mutation_probability': 0.1,
'elit_ratio': 0.01,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'uniform',
'max_iteration_without_improv': None}
# model = ga(function=f, dimension=ln, variable_type='real', variable_boundaries=varbound)
model = ga(function=f, dimension=ln, variable_type='real', variable_boundaries=varbound,
algorithm_parameters=algorithm_param)
model.run()
def optimize(
graph: RoutingGraphDirected,
num_trucks: int,
num_routes: int,
time_weight: float,
environmental_weight: float,
distance_weight: float,
cargo_weight: float,
):
varbound = np.array([[0, len(graph.vertex_list) - 1]] * num_trucks *
num_routes)
model = ga(
function=lambda X: loss(
X,
num_trucks,
num_routes,
graph,
time_weight,
environmental_weight,
distance_weight,
cargo_weight,
),
dimension=num_trucks * num_routes,
variable_type="int",
variable_boundaries=varbound,
)
model.run()
def roda_genetic(a,b,k=0):
"""
Função proxy para a utilização de algoritmos genéticos para o cálculo do custo
para a transformação de uma string em outra
:param a: palavra 1
:param b: palavra 2
:return: custo de transformação da string 1 em string 2 segundo o algoritmo genético
"""
avalia = AvaliaMovimento(a,b)
interacoes = len(a)*len(b)
populacao_avaliada = max(math.factorial(len(a))/interacoes,2)
if k==0:
num_parametros = math.factorial(len(a))
else:
num_parametros=k
varbound = np.array([[0,3]]*num_parametros)
algorithm_param = {'max_num_iteration': interacoes,\
'population_size':populacao_avaliada,\
'mutation_probability':0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.7,\
'parents_portion': 0.3,\
'crossover_type':'uniform',\
'max_iteration_without_improv':max(interacoes,10)}
model = ga(function=avalia.custo,dimension=num_parametros,variable_type='int',variable_boundaries=varbound\
,convergence_curve=False,progress_bar=False,\
algorithm_parameters=algorithm_param)
model.run()
return model
def solve(self,
population=100,
mutation_probability=0.1,
crossover_probability=0.5,
crossover_type='uniform',
parents_portion=0.3,
max_iteration=None):
params = {
'max_num_iteration': max_iteration,
'population_size': population,
'mutation_probability': mutation_probability,
'elit_ratio': (1 / population),
'crossover_probability': crossover_probability,
'parents_portion': parents_portion,
'crossover_type': crossover_type,
'max_iteration_without_improv': None
}
def target(x):
penalty = 0
for edge in self.edges:
if x[edge[0] - 1] == x[edge[1] - 1]:
penalty += self.no_vertex
return np.unique(x).size + penalty
var_bounds = np.array([[1, self.no_vertex]] * self.no_vertex)
model = ga(function=target,
dimension=self.no_vertex,
variable_type='int',
variable_boundaries=var_bounds,
algorithm_parameters=params)
model.run()
return model.best_function, model.best_variable, model.report
def minimize_genetic_abs_diff(space_vals, model_vals, tolerance_mm,
normal_vectors):
(n, space_vals, space_vals_flat,
model_vals) = extract_optimization_vars(space_vals, model_vals)
a0 = extract_abs_diff_ratios(space_vals, model_vals)
# get bounds
varbound = np.around(
extract_bounds(space_vals_flat, tolerance_mm, for_ga=True)).astype(int)
# set up global vars used in ga function
global model_vals_ga, a0_ga, n_ga, space_vals_start_flat_ga, normal_vectors_ga
model_vals_ga = model_vals
a0_ga = a0
n_ga = n
space_vals_start_flat_ga = space_vals_flat
normal_vectors_ga = normal_vectors
# initialize and run model
model = ga(function=abs_diff_ratio_function_ga,
dimension=9,
variable_type='int',
variable_boundaries=varbound,
convergence_curve=False)
model.run()
return model.output_dict['variable'].reshape(n, 3).tolist()
def runGeneticAlgorithm(self):
# ga
varbound = np.array([[0, 100]] * len(self.palette))
algorithm_parameters={'max_num_iteration': 4000,\
'population_size':100,\
'mutation_probability':0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.5,\
'parents_portion': 0.3,\
'crossover_type':'uniform',\
'max_iteration_without_improv':None}
model = ga(function=self.f,
dimension=len(self.palette),
variable_type='real',
variable_boundaries=varbound,
convergence_curve=True,
algorithm_parameters=algorithm_parameters)
#print(model.param)
model.run()
print("output variable is: ")
output = np.matrix(model.output_dict['variable'])
output /= output.max()
print(output)
return output
def test_rastrigin_initialized():
parameters = {
'max_num_iteration': 100,
'population_size': 2000,
'mutation_probability': 0.1,
'elit_ratio': 0.02,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'two_point',
'max_iteration_without_improv': None,
'multiprocessing_ncpus': 4,
'multiprocessing_engine': None,
}
dimension = 200
varbound = np.array([[-5.12, 5.12]] * dimension)
model = ga(function=f,
dimension=dimension,
variable_type='real',
variable_boundaries=varbound,
algorithm_parameters=parameters)
model.run(plot=True,
initial_idv=np.random.random(size=dimension) * 5.12 - 10.24)
assert model.best_function < 1000
def genetic_alg(self, bounds, params, timeOut=10.0):
model = ga(
function=self.distance,
dimension=bounds.shape[0],
variable_type='real',
variable_boundaries=bounds,
algorithm_parameters=params,
function_timeout=timeOut
)
return model
def start(tickers_size,
tickers_pool,
prices_pool,
algorithm_param,
target='calmar'):
# varbound = np.array([[0.5, 1.5], [1, 100], [0, 1]])
varbound = np.array([[0, len(tickers_pool) - 1]] * tickers_size)
# remember to count down 1
# vartype = np.array([['real'], ['int'], ['int']])
vartype = np.array([['int'] * tickers_size])
if target == 'calmar':
model = ga(
function=f, # i cant submit a self.f to ga lib
dimension=tickers_size,
variable_type='int',
# variable_type_mixed = vartype
variable_boundaries=varbound,
algorithm_parameters=algorithm_param,
tickers_pool=tickers_pool,
prices_pool=prices_pool,
function_timeout=20)
else:
model = ga(
function=f_sharpe, # i cant submit a self.f to ga lib
dimension=tickers_size,
variable_type='int',
# variable_type_mixed = vartype
variable_boundaries=varbound,
algorithm_parameters=algorithm_param,
tickers_pool=tickers_pool,
prices_pool=prices_pool,
function_timeout=20)
model.run()
solution = model.output_dict
df = pd.DataFrame.from_dict(solution)
root_list = df['variable'].values.tolist()
return root_list
def genetic_vehicle(periods,current_charge,charge_speed,lambda_mix,lambda_cost, mix_list, cost_list, tweak):
def cost_function(X):
penalty = 0
charge = 0
count = 0
objective = current_charge
difference = 0
while(objective < 100):
objective += charge_speed * 30
difference += 1
for i in range(periods):
if(X[i] == 1):
count += 1
charge += 30 * X[i] * charge_speed
if(difference != count):
penalty += abs(difference - count) * 100
if(X[0] == 1 and tweak == 0):
penalty += 10
#if(current_charge + charge < 100):
# penalty += abs(count - objective)*0.1
#if(current_charge + charge > 120):
# penalty += abs(count - objective)*0.1
score = []
for i in range(periods):
score.append((lambda_mix * mix_list[i] - lambda_cost * cost_list[i]) * X[i])
return - (sum(score) / len(score)) + penalty
algorithm_param = {'max_num_iteration': 300,\
'population_size':150,\
'mutation_probability':0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.5,\
'parents_portion': 0.3,\
'crossover_type':'uniform',\
'max_iteration_without_improv':100,
}
model=ga(function=cost_function,dimension=periods,variable_type='bool',algorithm_parameters=algorithm_param,convergence_curve=False)
model.run()
solution=model.output_dict
return solution.get('variable')
def genetic_algo(nr_dict, pizzas):
"""
The matrix representing the solution of this problem is
on the column the pizza_id and on the row the group_id.
If an entry = true then that pizza_id will get delivered
to that group_id.
"""
val = Validation(nr_dict, pizzas)
dimensions = (nr_dict['t2'] + nr_dict['t3'] + nr_dict['t4']) * len(pizzas)
print(dimensions)
model = ga(function=val.validate,
dimension=dimensions,
variable_type='bool')
model.run()
def egg_ga(maxRun):
for i in range(maxRun):
global egg_fVals
global fVals
egg_fVals = []
egg_sol = ga(function=eggcrate,
dimension=2,
variable_type='real',
variable_boundaries=egg_bound,
algorithm_parameters=algorithm_param)
start_time = time.process_time()
egg_sol.run()
end_time = time.process_time()
print("Execution Time: ", end_time - start_time)
fVals.append(egg_fVals)
plot_gaRosen(maxRun)
return egg_sol
def main():
varBound = np.array([[0,500],[1,5],[0,1],[2,9],[0,100],[1,7],[0,100],[1,2],[0,6]])
algorithm_param = {'max_num_iteration': None,\
'population_size':250,\
'mutation_probability':0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.5,\
'parents_portion': 0.3,\
'crossover_type':'two_point',\
'max_iteration_without_improv':None,\
'func_timeout': 10000000}
model = ga(function=minRuntime,dimension=9,variable_type='int',variable_boundaries=varBound, algorithm_parameters=algorithm_param)
model.run()
def obtenerSolucion():
global listaTotal,noLeGustaOrdenado,maxNoGustaIngredientesIni,maxNoGustaIngredientesFin,profundidadIngredientesIni,profundidadIngredientesFin
algorithm_param = {'max_num_iteration': 15,\
'population_size':20,\
'mutation_probability':0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.5,\
'parents_portion': 0.3,\
'crossover_type':'uniform',\
'max_iteration_without_improv':3}
model=ga(function=scoreSolucion,dimension=len(noLeGustaOrdenado),variable_type='bool',algorithm_parameters=algorithm_param)
model.run()
mejorSol=model.output_dict["variable"]
print(mejorSol)
return mejorSol
def rosen_ga(maxRun):
for i in range(maxRun):
global rosen_fVals
global rfVals
rosen_fVals = []
rosen_sol = ga(function=rosenbrock,
dimension=2,
variable_type='real',
variable_boundaries=rosen_bound,
algorithm_parameters=algorithm_param)
start_time = time.process_time()
rosen_sol.run()
end_time = time.process_time()
print("Execution Time: ", end_time - start_time)
rfVals.append(rosen_fVals)
plot_gaRosen(maxRun)
return rosen_sol
def _optimize_segment(region, dates, initial, attributes, weekly):
"""Optimize parameters on segment using simulation.
Args:
region (str): Region to run simulation for.
dates (tuple (2) of datetime.datetime): Limits of dates.
initial (tuple (5)): Initial values for (S,E,I,R,D).
attributes (str, optional): Attributes used for optimization, 'I', 'R' or 'D'.
weekly (bool, optional): Use weekly time slots if True, otherwise daily.
"""
fixparams = [None, .2, None, None] #.0064]
def _obj(pars):
return posterior.posterior_objective(pars,
region=region,
fixparams=fixparams,
weekly=weekly,
dates=dates,
initial=initial,
attributes=attributes,
parI=(1, 1),
parR=(1, 1),
parD=(1, 1))
algorithm_param = {
'max_num_iteration': 500,
'population_size': 70,
'mutation_probability': .65,
'elit_ratio': .05,
'crossover_probability': .8,
'parents_portion': .5,
'crossover_type': 'uniform',
'max_iteration_without_improv': 40
}
varbound = np.array([[0, 1], [0.033, .5], [0, .1]])
model = ga(function=_obj,
dimension=sum([i is None for i in fixparams]),
variable_type='real',
variable_boundaries=varbound,
convergence_curve=False,
progress_bar=True,
algorithm_parameters=algorithm_param)
model.run()
# best params
params = posterior._parse_params(model.output_dict['variable'], fixparams)
return params
def fit_parameters(self, error_func):
gamma1 = (1 / 10 + 1 / 5) / 2
gamma2 = 0.02
gamma3 = 0.02
beta = 0.14
num_cumulative_positiv = np.sum(self._observations[:, 0])
print(f"num_cumulative_positiv {num_cumulative_positiv}")
print(self._observations[:, 0])
tau = 28 / num_cumulative_positiv
delta = 5 / 18
params = Parameters()
params.add('gamma1', value=gamma1, min=1 / 10, max=1 / 5)
params.add('gamma2', value=gamma2, min=0, max=5)
params.add('gamma3', value=gamma3, min=0, max=5)
params.add('beta', value=beta, min=0.01, max=0.5)
params.add('tau', value=tau, min=tau * 0.8, max=tau * 1.2)
params.add('delta', value=delta, min=0.6 * delta, max=1.4 * delta)
# result = minimize(self._plumb,
# params,
# args=(len(self._observations), error_func),
# method='leastsq')
# report_fit(result)
# self._fit_params = result.params
print(params.items())
bounds = np.array([(p.min, p.max) for p_name, p in params.items()])
self._error_func = error_func
gamodel = ga(function=self._plumb,
dimension=len(bounds),
variable_type='real',
variable_boundaries=bounds)
gamodel.run()
self._fit_params = self._params_array_to_dict(
gamodel.output_dict['variable'])
def runGA():
params = {'max_num_iteration': 500,\
'population_size': 50,\
'mutation_probability': 0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.5,\
'parents_portion': 0.3,\
'crossover_type': 'uniform',\
'max_iteration_without_improv': None}
model=ga(function = fitness,\
dimension = 21,\
variable_type = 'bool',\
variable_boundaries = None,\
algorithm_parameters = params)
print(model.param)
model.run()
def run_genetic_merge(items, template):
algorithm_param = {'max_num_iteration': 3000,
'population_size': 100,
'mutation_probability': 0.1,
'elit_ratio': 0.01,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'uniform',
'max_iteration_without_improv': 150}
f = partial(compute_volume_ga, items, template)
np.random.seed(0)
model = ga(function=f,
dimension=len(items),
variable_type='bool',
algorithm_parameters=algorithm_param)
model.run()
return model.best_variable.astype(dtype="bool")
def solve_with_GA(params):
# algorithm = AlgorithmGaBihc([params['function'],params['bounds'][0]],params['maxFes'],dimensions=params['dimensions'],steps=10000,popSize=100)
# try:
# _explored_points = algorithm.run()
# except Exception as _:
# pass
# return {'f_value': algorithm.be, 'solution': algorithm.bv}
alg = ga(function=params['function'],dimension=params['ndim'],variable_type='real',variable_boundaries=np.array(params['bounds']),algorithm_parameters={
'max_num_iteration': 100000,\
'population_size':100,\
'mutation_probability':0.1,\
'elit_ratio': 0.01,\
'crossover_probability': 0.5,\
'parents_portion': 0.3,\
'crossover_type':'uniform',\
'max_iteration_without_improv':None})
alg.run()
return {'f_value': alg.best_function, 'solution': alg.best_variable}
def getModel(self):
algoParam = {
'max_num_iteration': 40,
'population_size': 600,
'mutation_probability': 0.1,
'elit_ratio': 0,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'uniform',
'max_iteration_without_improv': None
}
gaModel = ga(function=self.objFunction,
dimension=self.space.dNum,
variable_type='real',
variable_boundaries=self.space.getAllDimensionBounds(),
convergence_curve=self.convergence_curve,
algorithm_parameters=algoParam,
progress_bar=self.progress_bar)
return gaModel
def fit_parameters_ga(self, error_func):
# Fit parameters using a genetic algorithm
params = self.get_initial_parameters()
bounds = np.array([(p.min, p.max) for p_name, p in params.items()])
self._error_func = error_func
gamodel = ga(function=self._plumb_ga,
dimension=len(bounds),
variable_type='real',
variable_boundaries=bounds)
gamodel.run()
self._fit_params = self._params_array_to_dict(
gamodel.output_dict['variable'])
for p_name, p in params.items():
print("{:10s} [{:.2f} - {:.2f}] : {:.2f}".format(
p_name, p.min, p.max, self._fit_params[p_name]))
def optimize_genetic(self):
print("optimizing via genetic alg...")
algorithm_param = {
'max_num_iteration': 1000,
'population_size': 100,
'mutation_probability': 0.1,
'elit_ratio': 0.01,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'uniform',
'max_iteration_without_improv': 100
}
bounds = self.algorithm.get_possible_parameters(self.algorithm)
model = ga(function=self.fitness,
dimension=len(bounds),
variable_type=type(bounds[0][0]).__name__,
variable_boundaries=numpy.array(bounds),
algorithm_parameters=algorithm_param)
val = model.run()
print(val)
def test_rastrigin():
parameters = {
'max_num_iteration': 1000,
'population_size': 200,
'mutation_probability': 0.1,
'elit_ratio': 0.02,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'two_point',
'max_iteration_without_improv': None,
'multiprocessing_ncpus': 4,
'multiprocessing_engine': None,
}
varbound = np.array([[-5.12, 5.12]] * 2)
model = ga(function=f,
dimension=2,
variable_type='real',
variable_boundaries=varbound,
algorithm_parameters=parameters)
model.run()
assert model.best_function < 1e-6
def genetic_lstm():
"""LSTM input/architecture genetic optimization"""
varbound = np.array([
[3, 64], #histo
[16, 64] #units
])
algorithm_parameters = {
'max_num_iteration': None,
'population_size': 100, #default is 100
'mutation_probability': 0.1,
'elit_ratio': 0.01,
'crossover_probability': 0.5,
'parents_portion': 0.3,
'crossover_type': 'uniform',
'max_iteration_without_improv': None
}
optim = ga(function=rmse_lstm,
dimension=2,
variable_type='int',
variable_boundaries=varbound,
function_timeout=60 * 60,
algorithm_parameters=algorithm_parameters)
optim.run()
def GA_model(algorithm_param):
"""
Simple genetic algorithm model based on the geneticalgorithm library
(https://pypi.org/project/geneticalgorithm/)
"""
def fitness_func(gene, render=False):
rewards = []
episodes = 3
for _ in range(episodes):
observation = env.reset()
for _ in range(goal_steps):
if render:
env.render()
# Perform logistic regression function here
# to get either a 0 (left) or 1 (right) action
weights = gene
a = sigmoid(observation.dot(weights.T))
action = int(to_action(a))
observation, reward, done, info = env.step(action)
rewards.append(reward)
if done:
break
result = -sum(rewards) / episodes
return result
dim = 4
varbound = np.array([[-1000, 1000]] * dim)
model = ga(function=fitness_func,
dimension=4,
variable_type='real',
variable_boundaries=varbound,
algorithm_parameters=algorithm_param)
model.run()
weights = model.output_dict['variable']
return weights
def genetic_algorithm(env: CompilerEnv):
def f(choices):
env.reset()
env.choices = choices = list(map(int, choices))
s = objective(env)
return s if s > 0 else float("inf")
model = ga(
function=f,
dimension=len(env.gcc_spec.options),
variable_type="int",
variable_boundaries=np.array([[-1,
min(FLAGS.max_range,
len(opt) - 1)]
for opt in env.gcc_spec.options]),
function_timeout=FLAGS.timeout,
algorithm_parameters={
"population_size":
FLAGS.pop_size,
"max_num_iteration":
max(1, int(FLAGS.gcc_search_budget / FLAGS.pop_size)),
"mutation_probability":
0.1,
"elit_ratio":
0.01,
"crossover_probability":
0.5,
"parents_portion":
0.3,
"crossover_type":
"uniform",
"max_iteration_without_improv":
None,
},
)
model.run()
return model.best_function