def plotResults(variable, data, templates, results):
resCan = Canvas()
leg = Legend(nTemplates + 2)
data.Draw('PE')
leg.AddEntry(data, style='LEP')
nBins = len(inputTemplates[variable]['data'][whichBinFromFile])
h_tSumAfter = Hist(nBins, 0, nBins, title='after_' + variable)
if useT1:
plotTemplateAfter(templates[tNames['t1']], results[tNames['t1']][0],
resCan, leg, h_tSumAfter)
pass
if useT2:
plotTemplateAfter(templates[tNames['t2']], results[tNames['t2']][0],
resCan, leg, h_tSumAfter)
pass
if useT3:
plotTemplateAfter(templates[tNames['t3']], results[tNames['t3']][0],
resCan, leg, h_tSumAfter)
pass
if useT4:
plotTemplateAfter(templates[tNames['t4']], results[tNames['t4']][0],
resCan, leg, h_tSumAfter)
pass
leg.Draw()
h_tSumAfter.SetLineColor(2)
h_tSumAfter.SetLineStyle(7)
h_tSumAfter.SetMarkerSize(0)
h_tSumAfter.Draw('SAME HIST')
resCan.Update()
return resCan, h_tSumAfter
def mpe_fitting(filename, run, num_photons, use_ideal=True):
run_number = int(run)
if filename[-5:] == '.root':
filename = filename[:-5]
s_data_path = './data/%s.p' % filename
if filename[:5] == 'nerix':
file_identifier = filename
else:
file_identifier = filename[-9:]
a_integral = pickle.load((open(s_data_path, 'r')))
# max_num_events used to limit amplitudes
max_num_events = len(a_integral)
s_path_to_save = './results/%s/' % (file_identifier)
l_colors = [4, 2, 8, 7, 5, 9] + [4, 2, 8, 7, 5, 9]
d_mpe_fit = {}
if file_identifier == '0062_0061':
d_mpe_fit['settings'] = [250, -1e6, 2e7]
d_mpe_fit['bkg_mean_low'] = -1e6
d_mpe_fit['bkg_mean_high'] = 2e6
d_mpe_fit['bkg_width_low'] = 1e4
d_mpe_fit['bkg_width_high'] = 1e6
d_mpe_fit['spe_mean_low'] = 3.5e6
d_mpe_fit['spe_mean_high'] = 1e7
d_mpe_fit['spe_width_low'] = 8e5
d_mpe_fit['spe_width_high'] = 3e6
d_mpe_fit['ua_mean_low'] = 1e5
d_mpe_fit['ua_mean_high'] = 5e6
d_mpe_fit['ua_width_low'] = 1e4
d_mpe_fit['ua_width_high'] = 1e6
d_mpe_fit['spe_mean_guess'] = 6e6
d_mpe_fit['spe_width_guess'] = 1.9e6
elif file_identifier == '0066_0065':
d_mpe_fit['settings'] = [250, -1e6, 1.2e7]
d_mpe_fit['bkg_mean_low'] = -1e6
d_mpe_fit['bkg_mean_high'] = 2e6
d_mpe_fit['bkg_width_low'] = 1e4
d_mpe_fit['bkg_width_high'] = 1e6
d_mpe_fit['spe_mean_low'] = 2e6
d_mpe_fit['spe_mean_high'] = 6e6
d_mpe_fit['spe_width_low'] = 8e5
d_mpe_fit['spe_width_high'] = 3e6
d_mpe_fit['ua_mean_low'] = 1e5
d_mpe_fit['ua_mean_high'] = 5e6
d_mpe_fit['ua_width_low'] = 1e4
d_mpe_fit['ua_width_high'] = 1e6
d_mpe_fit['spe_mean_guess'] = 3.6e6
d_mpe_fit['spe_width_guess'] = 1.1e6
elif file_identifier == '0067_0068':
d_mpe_fit['settings'] = [250, -1e6, 7.5e6]
d_mpe_fit['bkg_mean_low'] = -5e5
d_mpe_fit['bkg_mean_high'] = 5e5
d_mpe_fit['bkg_width_low'] = 1e4
d_mpe_fit['bkg_width_high'] = 1e6
d_mpe_fit['spe_mean_low'] = 1.2e6
d_mpe_fit['spe_mean_high'] = 3.5e6
d_mpe_fit['spe_width_low'] = 3e5
d_mpe_fit['spe_width_high'] = 1e6
d_mpe_fit['ua_mean_low'] = 1e5
d_mpe_fit['ua_mean_high'] = 1.5e6
d_mpe_fit['ua_width_low'] = 1e4
d_mpe_fit['ua_width_high'] = 1e6
d_mpe_fit['spe_mean_guess'] = 2.1e6
d_mpe_fit['spe_width_guess'] = 6.6e5
elif file_identifier == '0071_0072':
d_mpe_fit['settings'] = [250, -1e6, 3.4e7]
d_mpe_fit['bkg_mean_low'] = -1e6
d_mpe_fit['bkg_mean_high'] = 2e6
d_mpe_fit['bkg_width_low'] = 1e4
d_mpe_fit['bkg_width_high'] = 1e6
d_mpe_fit['spe_mean_low'] = 7.5e6
d_mpe_fit['spe_mean_high'] = 1.4e7
d_mpe_fit['spe_width_low'] = 1e6
d_mpe_fit['spe_width_high'] = 3.5e6
d_mpe_fit['ua_mean_low'] = 1e5
d_mpe_fit['ua_mean_high'] = 5e6
d_mpe_fit['ua_width_low'] = 1e4
d_mpe_fit['ua_width_high'] = 1e6
d_mpe_fit['spe_mean_guess'] = 9.5e6
d_mpe_fit['spe_width_guess'] = 2.9e6
elif file_identifier == '0073_0074':
d_mpe_fit['settings'] = [50, -1e6, 4.2e7]
d_mpe_fit['bkg_mean_low'] = -1e6
d_mpe_fit['bkg_mean_high'] = 2e6
d_mpe_fit['bkg_width_low'] = 1e5
d_mpe_fit['bkg_width_high'] = 2e6
d_mpe_fit['spe_mean_low'] = 8.5e6
d_mpe_fit['spe_mean_high'] = 1.4e7
d_mpe_fit['spe_width_low'] = 1.5e6
d_mpe_fit['spe_width_high'] = 4.5e6
d_mpe_fit['ua_mean_low'] = 5e5
d_mpe_fit['ua_mean_high'] = 4e6
d_mpe_fit['ua_width_low'] = 0.5e6
d_mpe_fit['ua_width_high'] = 1e6
d_mpe_fit['spe_mean_guess'] = 9.5e6
d_mpe_fit['spe_width_guess'] = 2.9e6
elif file_identifier == 'nerix_160418_1523':
d_mpe_fit['settings'] = [50, -5e5, 3.e6]
d_mpe_fit['bkg_mean_low'] = -5e5
d_mpe_fit['bkg_mean_high'] = 5e5
d_mpe_fit['bkg_width_low'] = 1e4
d_mpe_fit['bkg_width_high'] = 5e5
d_mpe_fit['spe_mean_low'] = 6e5
d_mpe_fit['spe_mean_high'] = 11e5
d_mpe_fit['spe_width_low'] = 3e5
d_mpe_fit['spe_width_high'] = 9e5
d_mpe_fit['ua_mean_low'] = 1e3
d_mpe_fit['ua_mean_high'] = 5e5
d_mpe_fit['ua_width_low'] = 1e4
d_mpe_fit['ua_width_high'] = 5e5
elif file_identifier == 'nerix_160418_1531':
d_mpe_fit['settings'] = [50, -5e5, 4.e6]
d_mpe_fit['bkg_mean_low'] = -5e5
d_mpe_fit['bkg_mean_high'] = 5e5
d_mpe_fit['bkg_width_low'] = 1e4
d_mpe_fit['bkg_width_high'] = 5e5
d_mpe_fit['spe_mean_low'] = 6e5
d_mpe_fit['spe_mean_high'] = 11e5
d_mpe_fit['spe_width_low'] = 3e5
d_mpe_fit['spe_width_high'] = 9e5
d_mpe_fit['ua_mean_low'] = 1e3
d_mpe_fit['ua_mean_high'] = 5e5
d_mpe_fit['ua_width_low'] = 1e4
d_mpe_fit['ua_width_high'] = 5e5
else:
print '\n\nSettings do not exist for given setup: %s\n\n' % (file_identifier)
sys.exit()
l_plots = ['plots', file_identifier]
par_names = ['p0_ampl', 'mean_bkg', 'width_bkg', 'mean_spe', 'width_spe'] + ['p%d_ampl' % (i + 1) for i in xrange(num_photons)]
a_integral = pickle.load((open(s_data_path, 'r')))
if use_ideal:
l_mpe_fit_func = ['[5]/(2*3.14*%d*[4]**2.)**0.5*TMath::Poisson(%d, [0])*exp(-0.5/%.1f*((x - %.1f*[3])/[4])**2)' % (iElectron, iElectron, iElectron, iElectron) for iElectron in xrange(1, num_photons + 1)]
else:
l_mpe_fit_func = ['[5]/(2*3.14*([2]**2. + %d*[4]**2.))**0.5*TMath::Poisson(%d, [0])*exp(-0.5*((x - %.1f*[3] - [1])/(%.1f*[4]**2 + [2]**2)**0.5)**2)' % (iElectron, iElectron, iElectron, iElectron) for iElectron in xrange(1, num_photons + 1)]
h_mpe_spec = Hist(*d_mpe_fit['settings'], name='h_mpe_spec', title='MPE Spectrum with Gaussian Fit - %s' % filename)
h_mpe_spec.SetMarkerSize(0)
h_mpe_spec.fill_array(a_integral)
c1 = Canvas()
h_mpe_spec.Draw()
s_bkg = '[5]/(2*3.14*[2]**2.)**0.5*TMath::Poisson(0, [0])*exp(-0.5*((x - [1])/[2])**2)'
s_under_amplified = '[6]/(2*3.14*[8]**2.)**0.5*exp(-0.5*((x - [7])/[8])**2)'
s_fit_mpe = '(%s) + (%s) + (%s)' % (s_bkg, s_under_amplified, ' + '.join(l_mpe_fit_func))
s_fit_mpe = '(%s)*([1] < [3] ? 1. : 0.)*([1] < [7] ? 1. : 0.)*([7] < [3] ? 1. : 0.)' % (s_fit_mpe)
fit_mpe = root.TF1('fit_mpe', s_fit_mpe, *d_mpe_fit['settings'][1:])
fit_mpe.SetLineColor(46)
fit_mpe.SetLineStyle(2)
fit_mpe.SetLineWidth(3)
h_mpe_spec.GetXaxis().SetTitle('Integrated Charge [e-]')
h_mpe_spec.GetYaxis().SetTitle('Counts')
h_mpe_spec.GetYaxis().SetTitleOffset(1.4)
h_mpe_spec.SetStats(0)
c1.SetLogy()
fit_mpe.SetParLimits(0, 0.9, 2.5)
fit_mpe.SetParameter(0, 1.3)
fit_mpe.SetParLimits(1, d_mpe_fit['bkg_mean_low'], d_mpe_fit['bkg_mean_high'])
fit_mpe.SetParameter(1, (d_mpe_fit['bkg_mean_low']+d_mpe_fit['bkg_mean_high'])/2.)
fit_mpe.SetParLimits(2, d_mpe_fit['bkg_width_low'], d_mpe_fit['bkg_width_high'])
fit_mpe.SetParameter(2, (d_mpe_fit['bkg_width_low']+d_mpe_fit['bkg_width_high'])/2.)
fit_mpe.SetParLimits(3, d_mpe_fit['spe_mean_low'], d_mpe_fit['spe_mean_high'])
fit_mpe.SetParameter(3, d_mpe_fit['spe_mean_guess'])
fit_mpe.SetParLimits(4, d_mpe_fit['spe_width_low'], d_mpe_fit['spe_width_high'])
fit_mpe.SetParameter(4, d_mpe_fit['spe_width_guess'])
fit_mpe.SetParLimits(5, 10, max_num_events*1e6)
fit_mpe.SetParameter(5, (10+max_num_events*1e6)/2.)
fit_mpe.SetParLimits(6, 10, max_num_events*1e6)
fit_mpe.SetParameter(6, (10+max_num_events*1e6)/2.)
fit_mpe.SetParLimits(7, d_mpe_fit['ua_mean_low'], d_mpe_fit['ua_mean_high'])
fit_mpe.SetParameter(7, (d_mpe_fit['ua_mean_low']+d_mpe_fit['ua_mean_high'])/2.)
fit_mpe.SetParLimits(8, d_mpe_fit['ua_width_low'], d_mpe_fit['ua_width_high'])
fit_mpe.SetParameter(8, (d_mpe_fit['ua_width_low']+d_mpe_fit['ua_width_high'])/2.)
"""
for i, guess in enumerate(mpe_par_guesses):
fit_mpe.SetParameter(i, guess)
for i in xrange(len(par_names)):
fit_mpe.SetParName(i, par_names[i])
for photon in xrange(num_photons):
fit_mpe.SetParLimits(5 + photon, 0, max_num_events)
"""
fitResult = h_mpe_spec.Fit('fit_mpe', 'MILES')
# draw individual peaks
s_gaussian = '[0]*exp(-0.5/%.1f*((x - %.1f*[1])/[2])**2)'
l_functions = []
l_individual_integrals = [0. for i in xrange(num_photons+2)]
for i in xrange(num_photons + 2):
l_functions.append(root.TF1('peak_%d' % i, '[0]*exp(-0.5*((x - [1])/[2])**2)', *d_mpe_fit['settings'][1:]))
# set parameters
if i == 0:
ampl = fit_mpe.GetParameter(5)*root.TMath.Poisson(0, fit_mpe.GetParameter(0))
mean = fit_mpe.GetParameter(1)
width = fit_mpe.GetParameter(2)
if width > 0:
ampl /= (2*3.14*width**2.)**0.5
l_functions[i].SetParameters(ampl, mean, width)
# under amplified peak
elif i == (num_photons + 1):
ampl = fit_mpe.GetParameter(6)
mean = fit_mpe.GetParameter(7)
width = fit_mpe.GetParameter(8)
if width > 0:
ampl /= (2*3.14*width**2.)**0.5
l_functions[i].SetParameters(ampl, mean, width)
else:
ampl = fit_mpe.GetParameter(5)*root.TMath.Poisson(i, fit_mpe.GetParameter(0))
if use_ideal:
mean = fit_mpe.GetParameter(3) * i
width = fit_mpe.GetParameter(4)*i**0.5
else:
mean = fit_mpe.GetParameter(3)*i + fit_mpe.GetParameter(1)
width = (fit_mpe.GetParameter(4)**2*i + fit_mpe.GetParameter(2)**2)**0.5
if width > 0:
ampl /= (2*3.14*width**2.)**0.5
l_functions[i].SetParameters(ampl, mean, width)
l_individual_integrals[i] = ampl*width*(2*3.1415)**0.5
l_functions[i].SetLineColor(l_colors[i])
l_functions[i].Draw('same')
c1.Update()
fitStatus = fitResult.CovMatrixStatus()
if fitStatus != 3:
neriX_analysis.failure_message('Fit failed, please adjust guesses and try again.')
fit_successful = False
else:
neriX_analysis.success_message('Fit successful, please copy output to appropriate files.')
fit_successful = True
#if not os.path.exists(sPathToSaveOutput):
# os.makedirs(sPathToSaveOutput)
fitter = root.TVirtualFitter.Fitter(fit_mpe)
#fitter = root.TVirtualFitter.GetFitter()
amin = np.asarray([0], dtype=np.float64)
dum1 = np.asarray([0], dtype=np.float64)
dum2 = np.asarray([0], dtype=np.float64)
dum3 = np.asarray([0], dtype=np.int32)
dum4 = np.asarray([0], dtype=np.int32)
fitter.GetStats(amin, dum1, dum2, dum3, dum4)
print '\n\namin for %d photons: %f' % (num_photons, amin)
print 'fAmin for %d photons: %f\n\n' % (num_photons, root.gMinuit.fAmin)
print fit_mpe.GetChisquare()
# draw tpavetext
tpt_mpe = root.TPaveText(.55,.75,.85,.85,'blNDC')
tpt_mpe.AddText('#mu_{SPE} = %.2e #pm %.2e' % (fit_mpe.GetParameter(3), fit_mpe.GetParError(3)))
tpt_mpe.AddText('#sigma_{SPE} = %.2e #pm %.2e' % (fit_mpe.GetParameter(4), fit_mpe.GetParError(4)))
tpt_mpe.Draw('same')
tpt_mpe.SetTextColor(root.kBlack)
tpt_mpe.SetFillStyle(0)
tpt_mpe.SetBorderSize(0)
c1.Update()
neriX_analysis.save_plot(l_plots, c1, 'mpe_poisson_gaussian_fit_%s' % (file_identifier))
return (0,0,0)
def makeDiscr(train_discr_dict, discr_dict, outfile, xtitle="discriminator"):
c = ROOT.TCanvas("c", "c", 800, 500)
ROOT.gStyle.SetOptStat(0)
ROOT.gPad.SetMargin(0.15, 0.1, 0.2, 0.1)
#ROOT.gPad.SetLogy(1)
#ROOT.gPad.SetGrid(1,1)
ROOT.gStyle.SetGridColor(17)
l = TLegend(0.17, 0.75, 0.88, 0.88)
l.SetTextSize(0.055)
l.SetBorderSize(0)
l.SetFillStyle(0)
l.SetNColumns(2)
colors = [2, 1, 4, ROOT.kCyan + 2]
counter = 0
for leg, discr in train_discr_dict.iteritems():
a = Hist(30, 0, 1)
#fill_hist_with_ndarray(a, discr)
a.fill_array(discr)
a.SetLineColor(colors[counter])
a.SetLineWidth(2)
a.GetXaxis().SetTitle(xtitle)
a.GetXaxis().SetLabelSize(0.05)
a.GetXaxis().SetTitleSize(0.06)
a.GetXaxis().SetTitleOffset(1.45)
a.GetYaxis().SetTitle("a.u.")
a.GetYaxis().SetTickSize(0)
a.GetYaxis().SetLabelSize(0)
a.GetYaxis().SetTitleSize(0.06)
a.GetYaxis().SetTitleOffset(0.9)
a.Scale(1. / a.Integral())
#a.GetYaxis().SetRangeUser(0.00001,100)
a.GetYaxis().SetRangeUser(0, 0.9)
if counter == 0: a.draw("hist")
else: a.draw("same hist")
l.AddEntry(a, leg, "l")
counter += 1
counter = 0
for leg, discr in discr_dict.iteritems():
a = Hist(30, 0, 1)
#fill_hist_with_ndarray(a, discr)
a.fill_array(discr)
a.SetLineColor(colors[counter])
a.SetMarkerColor(colors[counter])
a.SetMarkerStyle(34)
a.SetMarkerSize(1.8)
a.SetLineWidth(2)
a.GetXaxis().SetTitle(xtitle)
a.GetXaxis().SetLabelSize(0.05)
a.GetXaxis().SetTitleSize(0.06)
a.GetXaxis().SetTitleOffset(1.45)
a.GetYaxis().SetTitle("a.u.")
a.GetYaxis().SetTickSize(0)
a.GetYaxis().SetLabelSize(0)
a.GetYaxis().SetTitleSize(0.06)
a.GetYaxis().SetTitleOffset(0.9)
a.Scale(1. / a.Integral())
#a.GetYaxis().SetRangeUser(0.00001,100)
a.GetYaxis().SetRangeUser(0, 0.4)
a.draw("same p X0")
l.AddEntry(a, leg, "p")
counter += 1
# counter = 0
# for leg,discr in train_discr_dict.iteritems():
# d = Hist(30, 0, 1)
# d.fill_array(discr)
# d.SetLineColor(colors[counter])
# d.SetLineWidth(2)
# l.AddEntry(d,leg,"l")
#
# b = Hist(30, 0, 1)
# d.fill_array(discr_dict[leg.split(" ")[0] + " test"])
# b.SetLineColor(colors[counter])
# b.SetMarkerColor(colors[counter])
# b.SetMarkerStyle(34)
# b.SetMarkerSize(1.8)
# b.SetLineWidth(2)
# l.AddEntry(b,leg,"p")
# counter += 1
l.Draw("same")
c.SaveAs(outfile)
def plot_difference(difference, centre_of_mass, channel, variable, k_value,
tau_value, output_folder, output_formats):
stats = len(difference)
values, errors = [], []
add_value = values.append
add_error = errors.append
for value, error in difference:
add_value(value)
add_error(error)
min_x, max_x = min(values), max(values)
abs_max = int(max(abs(min_x), max_x))
# n_x_bins = 2 * abs_max * 10
h_values = Hist(100, -abs_max, abs_max)
fill_value = h_values.Fill
for value in values:
fill_value(value)
plt.figure(figsize=(16, 16), dpi=200, facecolor='white')
axes = plt.axes()
h_values.SetMarkerSize(CMS.data_marker_size)
rplt.errorbar(h_values, xerr=True, emptybins=True, axes=axes)
channel_label = latex_labels.channel_latex[channel]
var_label = latex_labels.variables_latex[variable]
title_template = 'SVD unfolding performance for unfolding of {variable}\n'
title_template += '$\sqrt{{s}}$ = {com} TeV, {channel}, {value}'
title = title_template.format(variable=var_label,
com=centre_of_mass,
channel=channel_label,
value=get_value_title(k_value, tau_value))
plt.xlabel('$\mathrm{unfolded} - \mathrm{true}$', CMS.x_axis_title)
plt.ylabel('number of toy experiments', CMS.y_axis_title)
plt.tick_params(**CMS.axis_label_major)
plt.tick_params(**CMS.axis_label_minor)
plt.title(title, CMS.title)
plt.tight_layout()
for save in output_formats:
plt.savefig(output_folder + 'difference_stats_' + str(stats) + '.' +
save)
min_x, max_x = min(errors), max(errors)
h_errors = Hist(1000, min_x, max_x)
fill_value = h_errors.Fill
for error in errors:
fill_value(error)
plt.figure(figsize=(16, 16), dpi=200, facecolor='white')
axes = plt.axes()
h_errors.SetMarkerSize(CMS.data_marker_size)
rplt.errorbar(h_errors, xerr=True, emptybins=True, axes=axes)
plt.xlabel('$\sigma(\mathrm{unfolded} - \mathrm{true})$', CMS.x_axis_title)
plt.ylabel('number of toy experiments', CMS.y_axis_title)
plt.tick_params(**CMS.axis_label_major)
plt.tick_params(**CMS.axis_label_minor)
plt.title(title_template, CMS.title)
plt.tight_layout()
for save in output_formats:
plt.savefig(output_folder + 'difference_errors_stats_' + str(stats) +
'.' + save)
