-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlikelihood_tau_discriminator_ptweighting.py
273 lines (249 loc) · 14.4 KB
/
likelihood_tau_discriminator_ptweighting.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
import ROOT
import math
import copy
import numpy as np
from CMGTools.H2TauTau.proto.plotter.ROCPlotter import *
ROOT.gROOT.SetBatch(True) # don't disply canvas
create_maps = False
create_discriminator = True
create_plots = True
#settings for track maps
nbins_pt = 22
nbins_dz = 20
nbins_dr = 20
pt_binning = np.array([0.0, 1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.2] + [1.0+39.0*0.7**x for x in reversed(range(10))])
dz_binning = np.array([0.4/nbins_dz*x for x in range(nbins_dz+1)])
dr_binning = np.array([0.5/nbins_dr*x for x in range(nbins_dr+1)])
#settings for discriminator hists
upperlim = 20.0
nbins_discrim = 200
file0 = ROOT.TFile("/eos/user/s/swozniew/TauPOG_files/ntuples/DYJetsToLL_M50_LO.root")
tree = file0.Get("tree")
c1 = ROOT.TCanvas()
if create_maps:
file1 = ROOT.TFile("track_maps.root", "RECREATE")
for charge in [["P", 1.0], ["M", -1.0]]:
for PUwin in [[0, 1], [30, 35], [60, 70]]:
print "Creating maps for PU {down} to {up} and charge {ch} ...".format(down=PUwin[0], up=PUwin[1], ch=charge[0])
#count event yields
n_sig = ROOT.TH1I("n_sig", "n_sig", 1, 0, 2)
tree.Draw("1>>n_sig", "tau_gen_match==5&&n_true_interactions>={down}&&n_true_interactions<={up}&&tau_charge=={ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[1]))
n_bkg = ROOT.TH1I("n_bkg", "n_bkg", 1, 0, 2)
tree.Draw("1>>n_bkg", "tau_gen_match==6&&gen_jet_pt>18&&abs(gen_jet_eta)<2.3&&n_true_interactions>={down}&&n_true_interactions<={up}&&tau_charge=={ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[1]))
print "Number of signal events: %i"%n_sig.GetBinContent(1)
print "Number of background events: %i"%n_bkg.GetBinContent(1)
#fill maps
histsig = ROOT.TH3D("sig_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0]), "sig_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0]), nbins_pt, pt_binning, nbins_dz, dz_binning, nbins_dr, dr_binning)
histbkg = ROOT.TH3D("bkg_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0]), "bkg_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0]), nbins_pt, pt_binning, nbins_dz, dz_binning, nbins_dr, dr_binning)
tree.Draw("tau_iso_ch_dr:abs(tau_iso_ch_dz):tau_iso_ch_pt>>sig_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0]), "(tau_gen_match==5&&n_true_interactions>={down}&&n_true_interactions<={up}&&tau_charge=={ch})".format(down=PUwin[0], up=PUwin[1], ch=charge[1]))
tree.Draw("tau_iso_ch_dr:abs(tau_iso_ch_dz):tau_iso_ch_pt>>bkg_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0]), "(tau_gen_match==6&&gen_jet_pt>18&&abs(gen_jet_eta)<2.3&&n_true_interactions>={down}&&n_true_interactions<={up}&&tau_charge=={ch})".format(down=PUwin[0], up=PUwin[1], ch=charge[1]))
#normalize maps
histsig.Scale(1.0/n_sig.GetBinContent(1))
histbkg.Scale(1.0/n_bkg.GetBinContent(1))
print "Expected integral number of charged tracks in the regarded phase space for real taus: %f"%histsig.Integral()
print "Expected integral number of charged tracks in the regarded phase space for fake taus: %f"%histbkg.Integral()
#save maps
histsig.Write()
histbkg.Write()
file1.Close()
if create_discriminator:
nevents_sig = [61016.0, 476480.0, 143537.0, 60690.0, 471014.0, 142300.0] #number of events per PU class (determined with first part of script)
nevents_bkg = [46440.0, 368436.0, 113992.0, 43208.0, 345605.0, 107060.0]
#load likelihood density
file1 = ROOT.TFile("track_maps.root")
LDsig = []
LDbkg = []
for charge in ["P", "M"]:
for PUwin in [[0, 1], [30, 35], [60, 70]]:
LDsig.append(file1.Get("sig_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0])))
LDbkg.append(file1.Get("bkg_PU{down}to{up}_{ch}".format(down=PUwin[0], up=PUwin[1], ch=charge[0])))
#calculate the 0 track likelihoods
print "Determine log likelihood difference between signal and fake for an event without tracks ..."
Ldelta_0 = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
for i in range(1, LDsig[0].GetNbinsX()+1):
for j in range(1, LDsig[0].GetNbinsY()+1):
for k in range(1, LDsig[0].GetNbinsZ()+1):
bin_i = LDsig[0].GetBin(i, j, k)
for l in range(6):
Ldelta_0[l] += math.log(1.0-LDbkg[l].GetBinContent(bin_i)) - math.log(1.0-LDsig[l].GetBinContent(bin_i))
print "... Positive charge: PU 0 to 1: %f; PU 30 to 35: %f; PU 60 to 70 %f --- Negative charge: PU 0 to 1: %f; PU 30 to 35: %f; PU 60 to 70 %f"%tuple(Ldelta_0)
#reset Ldelta_0
for i in range(3):
if Ldelta_0[i] > Ldelta_0[i+3]:
Ldelta_0[i] = Ldelta_0[i] - Ldelta_0[i+3]
Ldelta_0[i+3] = 0.0
else:
Ldelta_0[i+3] = Ldelta_0[i+3] - Ldelta_0[i]
Ldelta_0[i] = 0.0
#calculate iso and fill histograms
file2 = ROOT.TFile("LLR_discrim_hists_ptweighting.root", "RECREATE")
discrim_sig1 = ROOT.TH1I("RW_sig_PU0to1", "RW_sig_PU0to1", nbins_discrim, 0.0, upperlim)
discrim_sig2 = ROOT.TH1I("RW_sig_PU30to35", "RW_sig_PU30to35", nbins_discrim, 0.0, upperlim)
discrim_sig3 = ROOT.TH1I("RW_sig_PU60to70", "RW_sig_PU60to70", nbins_discrim, 0.0, upperlim)
discrim_sig = [discrim_sig1, discrim_sig2, discrim_sig3]
discrim_bkg1 = ROOT.TH1I("RW_bkg_PU0to1", "RW_bkg_PU0to1", nbins_discrim, 0.0, upperlim)
discrim_bkg2 = ROOT.TH1I("RW_bkg_PU30to35", "RW_bkg_PU30to35", nbins_discrim, 0.0, upperlim)
discrim_bkg3 = ROOT.TH1I("RW_bkg_PU60to70", "RW_bkg_PU60to70", nbins_discrim, 0.0, upperlim)
discrim_bkg = [discrim_bkg1, discrim_bkg2, discrim_bkg3]
nevents = tree.GetEntries()
print "Process %i events..."%nevents
i=0
for event in tree:
#if i==100000:
# break
i+=1
if (100.0*i/nevents)%10-(100.0*(i-1)/nevents)%10 < 0.0:
print "Processed %i0%%"%int(10*i/nevents)
is_true_tau = False
pu_class = -1
is_positive = True
#classify true tau or fake
if event.tau_gen_match==5:
is_true_tau = True
elif event.tau_gen_match==6 and event.gen_jet_pt>18 and abs(event.gen_jet_eta)<2.3:
is_true_tau = False
else:
continue
#classify PU
if event.n_true_interactions>=0 and event.n_true_interactions<=1:
pu_class = 0
elif event.n_true_interactions>=30 and event.n_true_interactions<=35:
pu_class = 1
elif event.n_true_interactions>=60 and event.n_true_interactions<=70:
pu_class = 2
else:
continue
#classify charge
if event.tau_charge==1.0:
is_positive = True
pu_class += 3 #evaluate with map created with oppositely charged taus
elif event.tau_charge==-1.0:
is_positive = False
else:
continue
#read charged tracks and determine bin numbers
ch_pt = event.tau_iso_ch_pt
ch_dz = event.tau_iso_ch_dz
ch_dr = event.tau_iso_ch_dr
track_bins = []
for track in zip(ch_pt, ch_dz, ch_dr):
track_bins.append([LDsig[pu_class].GetXaxis().FindBin(track[0]), LDsig[pu_class].GetYaxis().FindBin(abs(track[1])), LDsig[pu_class].GetZaxis().FindBin(track[2])]) #[int(track[0]/(20.0/nbins))+1, int(abs(track[1])/(0.4/nbins))+1, int(track[2]/(0.5/nbins))+1])
#calculate resulting log likelihood ratio
llr = 0.0 #Ldelta_0[pu_class]
for i, track in enumerate(track_bins):
exp_sig = LDsig[pu_class].GetBinContent(LDsig[pu_class].GetBin(*track))
exp_bkg = LDbkg[pu_class].GetBinContent(LDbkg[pu_class].GetBin(*track))
factor = 0.0 if exp_bkg == 0.0 else exp_sig / exp_bkg
#print "pt %f with factor %f"%(ch_pt[i], 1.0 - factor)
llr += ch_pt[i] * (1.0 - factor)
'''if exp_sig == 0.0 and exp_bkg == 0.0:
continue
elif exp_sig == 0.0:
#llr = 999.0
#break
exp_sig = 1.0/nevents_sig[pu_class]
#llr += math.log(exp_bkg) - math.log(1.0-exp_bkg) - math.log(exp_sig) + math.log(1.0-exp_sig)
elif exp_bkg == 0.0:
#llr = 0.0
#break
exp_bkg = 1.0/nevents_bkg[pu_class]
#else:
llr += math.log(exp_bkg) - math.log(1.0-exp_bkg) - math.log(exp_sig) + math.log(1.0-exp_sig)'''
if llr>upperlim*0.99:
llr=upperlim*0.99
if is_true_tau:
discrim_sig[pu_class%3].Fill(llr)
else:
discrim_bkg[pu_class%3].Fill(llr)
#create hists with standard isolation
histsig1 = ROOT.TH1F("sig_PU0to1", "sig_PU0to1", nbins_discrim, 0.0, 20.0)
histbkg1 = ROOT.TH1F("bkg_PU0to1", "bkg_PU0to1", nbins_discrim, 0.0, 20.0)
histsig2 = ROOT.TH1F("sig_PU30to35", "sig_PU30to35", nbins_discrim, 0.0, 20.0)
histbkg2 = ROOT.TH1F("bkg_PU30to35", "bkg_PU30to35", nbins_discrim, 0.0, 20.0)
histsig3 = ROOT.TH1F("sig_PU60to70", "sig_PU60to70", nbins_discrim, 0.0, 20.0)
histbkg3 = ROOT.TH1F("bkg_PU60to70", "bkg_PU60to70", nbins_discrim, 0.0, 20.0)
tree.Draw("(tau_chargedIsoPtSum*(tau_chargedIsoPtSum<20.0)+19.99*(tau_chargedIsoPtSum>=20.0))>>sig_PU0to1", "(tau_gen_match==5&&n_true_interactions>=0&&n_true_interactions<=1)")
tree.Draw("(tau_chargedIsoPtSum*(tau_chargedIsoPtSum<20.0)+19.99*(tau_chargedIsoPtSum>=20.0))>>bkg_PU0to1", "(tau_gen_match==6&&gen_jet_pt>18&&abs(gen_jet_eta)<2.3&&n_true_interactions>=0&&n_true_interactions<=1)")
tree.Draw("(tau_chargedIsoPtSum*(tau_chargedIsoPtSum<20.0)+19.99*(tau_chargedIsoPtSum>=20.0))>>sig_PU30to35", "(tau_gen_match==5&&n_true_interactions>=30&&n_true_interactions<=35)")
tree.Draw("(tau_chargedIsoPtSum*(tau_chargedIsoPtSum<20.0)+19.99*(tau_chargedIsoPtSum>=20.0))>>bkg_PU30to35", "(tau_gen_match==6&&gen_jet_pt>18&&abs(gen_jet_eta)<2.3&&n_true_interactions>=30&&n_true_interactions<=35)")
tree.Draw("(tau_chargedIsoPtSum*(tau_chargedIsoPtSum<20.0)+19.99*(tau_chargedIsoPtSum>=20.0))>>sig_PU60to70", "(tau_gen_match==5&&n_true_interactions>=60&&n_true_interactions<=70)")
tree.Draw("(tau_chargedIsoPtSum*(tau_chargedIsoPtSum<20.0)+19.99*(tau_chargedIsoPtSum>=20.0))>>bkg_PU60to70", "(tau_gen_match==6&&gen_jet_pt>18&&abs(gen_jet_eta)<2.3&&n_true_interactions>=60&&n_true_interactions<=70)")
histsig_std = [histsig1, histsig2, histsig3]
histbkg_std = [histbkg1, histbkg2, histbkg3]
for j in range(3):
file2.cd()
discrim_sig[j].Write()
discrim_bkg[j].Write()
histsig_std[j].Write()
histbkg_std[j].Write()
file2.Close()
file1.Close()
file0.Close()
if create_plots:
#load hists and create rebinned versions for plotting
rebinning_factor=5
PUs = ["PU0to1", "PU30to35", "PU60to70"]
file2 = ROOT.TFile("LLR_discrim_hists_ptweighting.root")
discrim_sig_load = []
discrim_bkg_load = []
stdiso_sig_load = []
stdiso_bkg_load = []
discrim_sig_rebinned = []
discrim_bkg_rebinned = []
for i in range(3):
discrim_sig_load.append(file2.Get("RW_sig_%s"%PUs[i]))
discrim_sig_rebinned.append(ROOT.TH1F("RW_sig_%s_rebinned"%PUs[i], "RW_sig_%s"%PUs[i], nbins_discrim/rebinning_factor, 0.0, upperlim))
discrim_sig_rebinned[i].SetBinContent(0, discrim_sig_load[i].GetBinContent(0))
discrim_sig_rebinned[i].SetBinContent(nbins_discrim/rebinning_factor+1, discrim_sig_load[i].GetBinContent(nbins_discrim+1))
discrim_bkg_load.append(file2.Get("RW_bkg_%s"%PUs[i]))
discrim_bkg_rebinned.append(ROOT.TH1F("RW_bkg_%s_rebinned"%PUs[i], "RW_sig_%s"%PUs[i], nbins_discrim/rebinning_factor, 0.0, upperlim))
discrim_bkg_rebinned[i].SetBinContent(0, discrim_sig_load[i].GetBinContent(0))
discrim_bkg_rebinned[i].SetBinContent(nbins_discrim/rebinning_factor+1, discrim_sig_load[i].GetBinContent(nbins_discrim+1))
for j in range(1, nbins_discrim/rebinning_factor+1):
content_sig = 0
content_bkg = 0
for k in range(rebinning_factor):
content_sig += discrim_sig_load[i].GetBinContent(j*rebinning_factor-k)
content_bkg += discrim_bkg_load[i].GetBinContent(j*rebinning_factor-k)
discrim_sig_rebinned[i].SetBinContent(j, content_sig)
discrim_bkg_rebinned[i].SetBinContent(j, content_bkg)
stdiso_sig_load.append(file2.Get("sig_%s"%PUs[i]))
stdiso_bkg_load.append(file2.Get("bkg_%s"%PUs[i]))
#create plots
colors = [1, 2, 4]
discrim_sig_rebinned[0].GetXaxis().SetTitle("Relative log likelihood ratio")
discrim_sig_rebinned[0].GetYaxis().SetTitle("Fraction of events")
discrim_sig_rebinned[0].Draw("hist") #for the range and labels
for i in range(3):
discrim_sig_rebinned[i].Scale(1.0/discrim_sig_rebinned[i].Integral())
discrim_bkg_rebinned[i].Scale(1.0/discrim_bkg_rebinned[i].Integral())
discrim_sig_rebinned[i].SetLineColor(colors[i])
discrim_bkg_rebinned[i].SetLineColor(colors[i])
discrim_bkg_rebinned[i].SetLineStyle(2)
discrim_sig_rebinned[i].SetLineWidth(3)
discrim_bkg_rebinned[i].SetLineWidth(3)
discrim_sig_rebinned[i].Draw("histsame")
discrim_bkg_rebinned[i].Draw("histsame")
legend1 = ROOT.TLegend(0.5, 0.76, 0.9, 0.9)
legend1.AddEntry(discrim_sig_rebinned[0], "real taus", "l")
legend1.AddEntry(discrim_bkg_rebinned[0], "fake taus", "l")
legend2 = ROOT.TLegend(0.5, 0.5, 0.9, 0.71)
legend2.AddEntry(discrim_sig_rebinned[0], "PU 0 to 1", "l")
legend2.AddEntry(discrim_sig_rebinned[1], "PU 30 to 35", "l")
legend2.AddEntry(discrim_sig_rebinned[2], "PU 60 to 70", "l")
legend1.Draw()
legend2.Draw()
c1.SaveAs("RW_lldiscrim.png")
rocs = []
for i in range(3):
rocs.append(histsToRoc(stdiso_sig_load[i], stdiso_bkg_load[i]))
rocs[-1].name = "standard iso %s"%PUs[i]
rocs[-1].title = "standard iso %s"%PUs[i]
for i in range(3):
rocs.append(histsToRoc(discrim_sig_load[i], discrim_bkg_load[i]))
rocs[-1].name = "RW iso %s"%PUs[i]
rocs[-1].title = "RW iso %s"%PUs[i]
#makeROCPlot(rocs, "RW_ROCs", xmin=0.9, logy=True)
#makeROCPlot(rocs, "RW_ROCs_fullRange", xmin=0.0, logy=True)
makeROCPlot(rocs, "RW_ROCs_WPs", ymax=0.2, logy=False)
file2.Close()