def generate(self):
x = np.copy(self.__input_img).flatten()
for i in range(self.__args.epochs):
self.__logger.info('Epoch: {} started'.format(i))
start_time = time.time()
x, min_val, info = optimizer(self.__loss_evaluator.loss, x,
fprime=self.__loss_evaluator.gradients,
maxfun=self.__args.max_fun,
disp=self.__args.verbose)
self.__logger.info('Loss for epoch {} has value: {}'.format(i, min_val))
# Save current generated image
img = x.copy().reshape((self.__input_img.shape[1], self.__input_img.shape[2], 3))
img = np.clip(img, 0, 255).astype(np.uint8)
input_filename = self.__get_file_name(self.__input_file)
style_filename = self.__get_file_name(self.__style_file)
output_filename = "{}/{}_{}_iteration_{}.png".format(self.__args.output_dir,
input_filename,
style_filename,
i)
scipy.misc.imsave(output_filename, img)
self.__logger.info("For epoch {} current image is saved in: ".format(i, output_filename))
end_time = time.time()
self.__logger.info('Iteration {} completed in {}secs'.format(i, round(end_time - start_time, 4)))
def main():
#Initialize parameters
#Rs = [0., 0.5, 1.0, 1.5, 0.0, 0., 0.5]
Rs = [2.5, 2.8, 3.0, 3.2, 3.4, 3.6, 3.8]
etas = [5, 10., 20., 40., 100., 200., 200.0]
parametervector = []
for i in range(len(Rs)):
parametervector.append(etas[i])
#parametervector.append(Rs[i])
#x0 = np.array(parametervector)
x0 = parametervector
optimizer = 'basinhopping'
#from scipy.optimize import minimize as optimizer
#optimizer_kwargs = {
# 'method' : 'BFGS',
# 'options': {'gtol': 1e-15, }
# }
if optimizer == 'BFGS':
from scipy.optimize import minimize as optimizer
optimizer_kwargs = {
'method': 'BFGS',
'options': {
'gtol': 1e-15,
}
}
#optimizer_kwargs = {'method':'BFGS', 'gtol': 1e-15, }
elif optimizer == 'basinhopping':
from scipy.optimize import basinhopping as optimizer
optimizer(get_loss, x0, niter=500, T=0.1, stepsize=0.01)
elif optimizer == 'L-BFGS-B':
from scipy.optimize import minimize as optimizer
optimizer_kwargs = {
'method': 'L-BFGS-B',
'options': {
'ftol': 1e-05,
'gtol': 1e-08,
'maxfun': 1000000,
'maxiter': 1000000
}
}
import scipy
from distutils.version import StrictVersion
if StrictVersion(scipy.__version__) >= StrictVersion('0.17.0'):
optimizer_kwargs['options']['maxls'] = 2000
def train(self, instances):
self.collect_x_y(instances, self.min_f, self.max_f)
initial_lambda = np.zeros(self.get_total_number())
returns = optimizer(self.calculate_log_likelihood, x0=initial_lambda,
fprime=self.calculate_gradient, iprint=1, maxiter=100)
self.final_lambdas = returns[0]
if pvalue < alpha:
print('Normality test failed, with {:.2e} < {}.'.format(pvalue, alpha))
print('Data is not Gaussian.')
print()
initial = [5300, 1.3, 0, 1000, 1, -4]
plt.hist(Data, bins=150, color='c', zorder=1, normed=True)
plt.plot(bin_centers,
combinedGaussian(bin_centers, *initial),
'r--',
zorder=10)
plt.title('Histogram with Initial Fit for Data')
plt.xlabel('Value')
plt.ylabel('Frequency')
plt.show()
plt.hist(Data, bins=150, color='c', zorder=1, normed=True)
params, cov = optimizer(combinedGaussian, bin_centers, n, p0=initial)
plt.plot(bin_centers, combinedGaussian(bin_centers, *params), 'r--', zorder=10)
plt.title('Fitted Combined Gaussian for Given Data Values')
plt.xlabel('Value')
plt.ylabel('Frequency')
plt.show()
param_names = ['C1 ', 'sigma1', 'mean1 ', 'C2 ', 'sigma2', 'mean2 ']
print("Fit parameters:")
for i in range(len(params)):
if (params[i] < 0):
print('{} = {:.3e}'.format(param_names[i], params[i]))
else:
print('{} = {:.3e}'.format(param_names[i], params[i]))
def main():
#Initialize parameters
#Rs = [0., 0.5, 1.0, 1.5, 0.0, 0., 0.5]
global eta
global Rs
#etas = [eta]
#Rss = [Rs]
etas = [100.]
Rss = [0.]
#g2_ref = calculate_fp([eta, Rs])
parametervector = []
for i in range(len(Rss)):
parametervector.append(etas[i])
parametervector.append(Rss[i])
#x0 = np.array(parametervector)
x0 = parametervector
optimizer = 'basinhopping'
#optimizer = 'L-BFGS-B'
#from scipy.optimize import minimize as optimizer
#optimizer_kwargs = {
# 'method' : 'BFGS',
# 'options': {'gtol': 1e-15, }
# }
if optimizer == 'BFGS':
from scipy.optimize import minimize as optimizer
optimizer_kwargs = {
'method' : 'BFGS',
'options': {'gtol': 1e-15, }
}
#optimizer_kwargs = {'method':'BFGS', 'gtol': 1e-15, }
elif optimizer == 'basinhopping':
from scipy.optimize import basinhopping as optimizer
minimizer_kwargs = {"method": "L-BFGS-B",
'options': {
'maxiter': 10,
#'maxfun':1
}
}
#mybounds = MyBounds()
results = optimizer(get_loss, x0,
#minimizer_kwargs=minimizer_kwargs,
niter=500,
T = 0.01,
stepsize = 1.0,
#niter_success=1000,
disp=True,
callback=print_fun,
# accept_test=mybounds
)
print results
sys.exit()
elif optimizer == 'L-BFGS-B':
from scipy.optimize import minimize as optimizer
optimizer_kwargs = {
'method': 'L-BFGS-B',
'options': {'ftol': 1e-05,
'gtol': 1e-18,
'maxfun': 1000000,
'maxiter':1500000,
'disp': True}
}
import scipy
from distutils.version import StrictVersion
if StrictVersion(scipy.__version__) >= StrictVersion('0.17.0'):
optimizer_kwargs['options']['maxls'] = 2000
optimizer(get_loss, x0, jac=False, **optimizer_kwargs)
return np.sum((y - Function(x, *fit_parameters))**2 / sigma**2)
filename = "C:/Users/ryank/Desktop/Work/Classes/Python/ASTR205/Data/"
filename += "ASTR 205 1-1.csv"
Data = np.loadtxt(filename, comments='#')
sigma = 1.9
plt.scatter(Data[:, 0], Data[:, 1], c='c', s=1)
plt.title('Initial Plot of the Given Data')
plt.xlabel('X values')
plt.ylabel('Y values')
plt.show()
guess = [1, -1, 1]
fit_params, fit_cov = optimizer(Function, Data[:, 0], Data[:, 1], p0=guess)
chi2 = chi_square(fit_params, Data[:, 0], Data[:, 1], sigma)
dof = len(Data[:, 0]) - len(fit_params)
print("\nGoodness of fit - chi square measure:")
print("Chi2/dof = {}\n".format(chi2 / dof))
fit_cov = fit_cov * dof / chi2
fit_params_error = np.sqrt(np.diag(fit_cov))
param_names = ['A', 'B', 'C']
print("Fit parameters:")
for i in range(len(fit_params)):
if (fit_params[i] < 0):
print('{} = {:.3e} +/- {:.3e}'.format(param_names[i], fit_params[i],
fit_params_error[i]))
else: