def get_mem_spectra_par_openmp(vq, trajectory, parameters):
test_frequency_range = np.array(parameters.frequency_range)
correlation_vector = []
progress_bar(0, "M.E. Method")
for i in range (vq.shape[1]):
correlation_vector.append(mem(test_frequency_range,
vq[:, i],
trajectory.get_time_step_average(),
coefficients=parameters.number_of_coefficients_mem))
progress_bar(float(i+1)/vq.shape[1],"M.E. Method")
correlation_vector = np.array(correlation_vector).T
return correlation_vector
def mem_coefficient_scan_analysis(vq, trajectory, parameters):
mem_full_dict = {}
for i in range(vq.shape[1]):
test_frequency_range = parameters.frequency_range
fit_data = []
scan_params = []
power_spectra = []
progress_bar(0, "M.E. Method")
for number_of_coefficients in parameters.mem_scan_range:
power_spectrum = mem(test_frequency_range,
vq[:, i],
trajectory.get_time_step_average(),
coefficients=number_of_coefficients)
height = np.max(power_spectrum)
position = test_frequency_range[np.argmax(power_spectrum)]
try:
fitParams, fitCovariances = curve_fit(lorentzian,
test_frequency_range,
power_spectrum,
p0=[position, 0.1, height, 0.0])
except:
print('Warning: Fitting error, skipping point {0}'.format(number_of_coefficients))
continue
maximum = fitParams[2]/(fitParams[1]*np.pi)
error = get_error_from_covariance(fitCovariances)/maximum
width = 2.0 * fitParams[1]
fit_data.append([number_of_coefficients, width, error])
scan_params.append(fitParams)
power_spectra.append(power_spectrum)
progress_bar(float(number_of_coefficients+1)/parameters.mem_scan_range[-1], "M.E. Method")
fit_data = np.array(fit_data).T
best_width = np.average(fit_data[1], weights=np.sqrt(1./fit_data[2]))
best_index = int(np.argmin(fit_data[2]))
power_spectrum = power_spectra[best_index]
mem_full_dict.update({i: [power_spectrum, best_width, best_index, fit_data, scan_params]})
for i in range(vq.shape[1]):
print "Peak # {0}".format(i+1)
print("------------------------------------")
print "Estimated width(FWHM): {0} THz".format(mem_full_dict[i][1])
fit_data = mem_full_dict[i][3]
scan_params = mem_full_dict[i][4]
best_index = mem_full_dict[i][2]
print "Position: {0} THz".format(scan_params[best_index][0])
print "Optimum coefficients num: {0}".format(fit_data[0][best_index])
print "Fitting Error: {0}".format(np.min(fit_data[2]))
print ("\n")
plt.figure(i+1)
plt.suptitle('Peak {0}'.format(i+1))
ax1 = plt.subplot2grid((2, 2), (0, 0), colspan=2)
ax2 = plt.subplot2grid((2, 2), (1, 0))
ax3 = plt.subplot2grid((2, 2), (1, 1))
ax1.set_xlabel('Number of coefficients')
ax1.set_ylabel('Width [THz]')
ax1.set_title('Peak width')
ax1.plot(fit_data[0], fit_data[1])
ax1.plot((fit_data[0][0], fit_data[0][-1]), (mem_full_dict[i][1], mem_full_dict[i][1]), 'k-')
ax2.set_xlabel('Number of coefficients')
ax2.set_ylabel('RMS^-1')
ax2.set_title('Fitting error/Max. (RMS)')
ax2.plot(fit_data[0], np.sqrt(1./fit_data[2]))
ax3.set_xlabel('Frequency [THz]')
ax3.set_title('Best curve fitting')
ax3.plot(test_frequency_range, mem_full_dict[i][0], label='Power spectrum')
ax3.plot(test_frequency_range,
lorentzian(test_frequency_range, *scan_params[best_index]),
label='Lorentzian fit')
plt.show()
