-
Notifications
You must be signed in to change notification settings - Fork 10
/
util.py
410 lines (374 loc) · 14.5 KB
/
util.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
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
import ROOT as r
import numpy as np
import os, sys, ast, type_func, cPickle
import pandas as pd
import h5py
from random import random
import itertools
r.gSystem.Load("libDelphes")
r.gInterpreter.Declare('#include "classes/DelphesClasses.h"')
r.gInterpreter.Declare('#include "external/ExRootAnalysis/ExRootTreeReader.h"')
r.gInterpreter.Declare('#include "external/ExRootAnalysis/ExRootResult.h"')
params = ['Px', 'Py', 'Pz', 'PT', 'E', 'D0', 'DZ', 'X', 'Y', 'Z']
max_len = 100
nothing = "[]"
jet_type_num = {0:'higgs', 1: 'top', 2: 'Z', 3: 'W+', 4: 'strange', 100: nothing}
jet_type = {v: k for k, v in jet_type_num.iteritems()}
POI = {6:'top', 25:'higgs', 23:'Z', 3: 'strange', 24:'W+'}
def print_attrs(object):
print "\n".join(dir(object))
def print_trefarray(array):
num_entries = array.GetEntries()
print "["
for i in range(num_entries):
print " ", str(array.At(i)) + ",\n",
print " ", array.At(num_entries - 1)
print "]"
def tref_array_to_numpy(array):
num_entries = array.GetEntries()
return np.array([array.At(i) for i in range(num_entries)])
def get_MOI(particle, branch_particle, count = 0, POI = POI):
if particle.M1 == 0 or branch_particle.At(particle.M1) == particle:
return nothing, count
if particle.PID in POI:
return POI[particle.PID], count
else:
try:
m1, c1 = get_MOI(branch_particle.At(particle.M1), branch_particle, count = count + 1)
except:
m1, c1 = nothing, 0
try:
m2, c2 = get_MOI(branch_particle.At(particle.M2), branch_particle, count = count + 1)
except:
m2, c2 = nothing, 0
if m1 == nothing:
return m2, c2
return m1, c1
def constituent_method(constituent, event, jet, moms):
s = str(constituent)
if "Muon" in s:
d = muon_to_dict(constituent, event, jet)
elif "Track" in s:
d = track_to_dict(constituent, event, jet)
elif "GenParticle" in s:
d = particle_to_dict(constituent, event, jet)
else:
return empty_dict(constituent, event, jet)
d['parents'] = moms[jet]
return d
def empty_dict(particle, event, jet):
return {}
def particle_to_dict(particle, event, jet,
params = ["Px", "Py", "Pz",
"PID", "E", "P", "T", "M1", "M2", "D1", "D2",
"D0", "DZ", "X", "Y", "Z",
"PT"]):
d = dict((param, getattr(particle, param)) for param in params)
d['event'] = event
d['jet'] = jet
d['track'] = False
d['muon'] = False
return d
def track_to_dict(track, event, jet, params = ["P", "PID"]):
p = track.Particle.GetObject()
if "GenParticle" in str(p):
d = particle_to_dict(p, event, jet)
else:
d = dict((param, getattr(track, param)) for param in params)
d['mom'] = nothing
d['event'] = event
d['jet'] = jet
d['muon'] = False
d['track'] = True
d['track_id'] = getattr(track, "PID")
return d
def muon_to_dict(muon, event, jet, params = ["PT"]):
p = muon.Particle.GetObject()
if not "0x0" in str(p):
d = particle_to_dict(p, event, jet)
else:
d = dict((param, getattr(track, param)) for param in params)
d['mom'] = nothing
d['event'] = event
d['jet'] = jet
d['track'] = False
d['muon'] = True
#print muon.Particle.GetObject()
return d
def get_mother_list(particle, branch_particle, mom_list = []):
m2 = particle.M2
m1 = particle.M1
if m1 >= branch_particle.GetEntries() or m2 >= branch_particle.GetEntries():
return sorted(list(set(mom_list)))
if particle == branch_particle.At(particle.M1):
return sorted(list(set(mom_list)))
if m1 == 0 and m2 == 0:
return sorted(list(set(mom_list)))
if m1 in mom_list and m2 in mom_list:
return sorted(list(set(mom_list)))
mom_list.insert(0, m1)
mom_list.insert(0, m2)
mom_list = list(set(mom_list))
new_list = get_mother_list(branch_particle.At(m1), branch_particle, mom_list) + \
get_mother_list(branch_particle.At(m2), branch_particle, mom_list)
return sorted(list(set(new_list)))
def not_tobject(index, branch_particle):
if "TObject" in str(type(branch_particle.At(index))):
return False
return True
def get_jet_parents(jet, branch_particle):
"""Get a fine-grained parent"""
#get list of parents
constituents = tref_array_to_numpy(jet.Constituents)
moms = [get_mother_list(particle, branch_particle) for particle in constituents]
#flatten list
moms = sorted(list(set(itertools.chain.from_iterable(moms))))
num_entries = branch_particle.GetEntries()
moms = [i for i in moms if i < num_entries]
pids = [branch_particle.At(i).PID for i in moms if not_tobject(i, branch_particle)]
jet_parents = []
a= jet_parents.append
if 25 in pids:
a('H')
if 6 in pids:
a('t')
if pids.count(5) >= 2:
a('b')
a('b')
if pids.count(5) == 1:
a('b')
if 24 in pids:
a('W+')
if pids.count(4) >= 2:
a('c')
a('c')
if pids.count(4) == 1:
a('c')
if pids.count(3) >= 2:
a('s')
a('s')
if pids.count(3) == 1:
a('s')
if 23 in pids:
a('Z')
jet_parents = sorted(jet_parents)
return str(jet_parents)
def assign_parent_list_to_dict(dic, parent):
dic['parents'] = parent
return dic
def printProgressBar (iteration, total, prefix = '', suffix = '', decimals = 1, length = 100, fill = '#'):
""" Call in a loop to create terminal progress bar
@params:
iteration - Required : current iteration (Int)
total - Required : total iterations (Int)
prefix - Optional : prefix string (Str)
suffix - Optional : suffix string (Str)
decimals - Optional : positive number of decimals in percent complete (Int)
length - Optional : character length of bar (Int)
fill - Optional : bar fill character (Str)
"""
percent = ("{0:." + str(decimals) + "f}").format(100 * (iteration / float(total)))
filledLength = int(length * iteration // total)
bar = fill * filledLength + '-' * (length - filledLength)
sys.stdout.write('\r%s |%s| %s%% %s' % (prefix, bar, percent, suffix))
sys.stdout.flush()
# Print New Line on Complete
if iteration == total:
print
def create_genjet_df(fname, do_one = False, filter_nothing = True):
chain = r.TChain("Delphes")
chain.Add(fname)
#want to read chain, including Delphes information
tree_reader = r.ExRootTreeReader(chain)
branch_genjet = tree_reader.UseBranch("GenJet")
#branch_fastjet = tree_reader.UseBranch("FastJet")
branch_particle = tree_reader.UseBranch("Particle")
branch_muon = tree_reader.UseBranch("Muon")
branch_eflow_track = tree_reader.UseBranch("EFlowTrack")
num_entries = tree_reader.GetEntries()
if do_one: #debugging
num_entries = 10
events = {}
id_counter = 0
for i in range(num_entries):
tree_reader.ReadEntry(i)
genjets = tref_array_to_numpy(branch_genjet)
moms = [get_jet_parents(jet, branch_particle) for jet in genjets]
events[i] = [constituent_method(c, i, j, moms) for j, jet in enumerate(genjets)
for c in tref_array_to_numpy(jet.Constituents)]
printProgressBar(i, num_entries)
print "Finished reading file..........%s" % fname
particle_dicts = [events[i][j]
for i in range(len(events))
for j in range(len(events[i]))
if events[i][j] != {}
]
df = pd.DataFrame.from_records(particle_dicts)
df = df_njet_index(df)
if filter_nothing:
print "Filtering out nothings..."
#df = df_filter_nothing(df)
df = df[df.parents != nothing]
df['count'] = df.groupby('njet')['parents'].transform('count')
return df
def df_filter_nothing(df):
dfs = filter_jet_list(df_get_jet_list(df))
return pd.concat(dfs)
def combine_dfs(dfs):
count = 0
for df in dfs:
df['njet'] += count
count += sorted(np.unique(df['njet'].values))[-1]
df = pd.concat(dfs)
return df
def create_genjet_combo_h5(fnames):
dfs = [create_genjet_h5(i) for i in fnames]
df = combine_dfs(dfs)
fname = fnames[0]
out = fname.split('.root')[0] + '_combo.h5'
print "Outputting to " + out
df.to_hdf(out, 'df', format = 'table')
print "h5 file written to " + out
return df
def create_genjet_h5(fname):
out = fname.split('.root')[0] + '.h5'
print "Generating dataframe..."
df = create_genjet_df(fname)
print "Outputting to " + out
df.to_hdf(out, 'df', format = 'table')
print "h5 file written to " + out
return df
def h5_to_target(fname, output = None, params = params, max_len = max_len):
df = h5_to_df(fname)
return df_to_target(df, output, params, max_len = max_len)
def h5_to_df(fname, jet_dict_file = None):
print "reading file"
df = pd.read_hdf(fname)
print "Generate dictionary"
types = sorted(list(set(df.parents.values)))
if jet_dict_file == None:
jet_dict = {}
for i in types:
#l = int(raw_input(i + ": "))
j = ast.literal_eval(i)
l = type_func.get_type(j)
print j, l
jet_dict[i] = l
cPickle.dump(jet_dict, open('jet_dict.pkl', 'wb'))
else:
jet_dict = cPickle.load(open(jet_dict_file, 'rb'))
print "Assigning jet_type..."
df['mom'] = df['parents'].map(jet_dict)
df['count'] = df.groupby('njet')['parents'].transform('count')
return df
def pad_values(vals, val = 0, max_len = max_len):
sr, sc = vals.shape
if max_len - sr > 0:
return np.pad(vals, ((0, max_len - sr), (0, 0)), mode='constant')
return vals[:max_len, :]
def print_accuracy( p, target ):
p_cat = np.argmax(p,axis=1)
test_target = np.argmax(target, axis = 1)
print "Fraction of good prediction"
print len(np.where( p_cat == test_target)[0])
print len(np.where( p_cat == test_target )[0])/float(len(p_cat)),"%"
def accuracy(p, target):
p_cat = np.argmax(p,axis=1)
test_target = np.argmax(target, axis = 1)
return len(np.where( p_cat == test_target)[0])/float(len(p_cat))
def df_njet_index(df):
"""Combine event # and jet #"""
events_jets = df[['event', 'jet']].values
njet = np.zeros(events_jets.shape[0])
prev = np.array([-1, -1])
count = -1
for ind, val in enumerate(events_jets):
if any(prev != val):
count += 1
prev = val
njet[ind] = count
df['njet'] = njet
return df
def df_get_jet_list(df):
"""Return a list of dataframes based on jet number"""
groups = df.groupby('njet')
print "Generated groupby object"
dfs = [groups.get_group(i) for i in groups.groups.keys()]
return dfs
def df_to_target(df, output = None, params = params, max_len = max_len):
df = df.sort_values(['njet', 'D0', 'DZ', 'PT'],ascending = False)
numpy_vals = df[['njet'] + params].values
moms = df[['njet', 'mom']].values
ma = max(moms[:, 1])
training = np.array([pad_values(i[:, 1:], max_len = max_len)
for i in np.split(numpy_vals, np.where(np.diff(numpy_vals[:,0]))[0]+1)])
training_target = np.array([get_list_from_num(i[0, 1], length = 1 + ma)
for i in np.split(moms, np.where(np.diff(moms[:, 0]))[0] + 1)])
return training, training_target
print "lookup"
dfs = df.groupby('njet')
jets = df.njet.unique()
jet_sub = np.random.choice(jets, 100, replace = False)
print "got jets"
training = np.array([pad_values(dfs.get_group(i)[params].sort_values(['D0', 'DZ', 'PT'],
ascending = False).values)
for i in jet_sub])
print "got training"
if output == None:
training_target = np.array([get_list_from_num(get_jet_num(dfs.get_group(i)))
for i in jet_sub])
else:
training_target = np.array([output for i in range(len(training))])
print "to numpy"
#training = np.array([pad_values(i) for i in training])
return training, training_target
def filter_jet_list(dfs):
"""Remove all the Nothings"""
dfs = [i for i in dfs if i.parents.values[0] !=nothing]
return dfs
def get_list_from_num(num, length = (len(jet_type_num) - 1)):
l = np.zeros(length)
l[int(num)] = 1
return l
def get_jet_num(df):
return df.mom.values[0]
def assign_jet_type(df, jet_dict):
"""Assign most appropriate jet type"""
jet_type_choice = jet_dict[df.parents.values[0]]
df['mom'] = jet_type_choice
return df
def make_test_split(training, target, test_size = 200):
"""Split training/target into training/target and test/target"""
print training.shape, target.shape
num = training.shape[0]
indices = np.random.choice(range(num), test_size, replace = False)
test = training[indices]
test_target = target[indices]
training = np.delete(training, indices, axis = 0)
target = np.delete(target, indices, axis = 0)
return training, target, test, test_target
def get_training_target_sample(training, target, sample_size):
indices = np.random.choice(range(training.shape[0]), sample_size, replace = False)
ntraining = training[indices]
ntarget = target[indices]
return ntraining, ntarget
def shuffle_together(training, target):
p = np.random.permutation(len(training))
return training[p], target[p]
def combine_sets(sets, sample_size = 10000):
"""Combine a list of (training, target, test, test_target) sets"""
a = min([i[0].shape[0] for i in sets])
sample_size = min(a, sample_size)
samples = [get_training_target_sample(i[0], i[1], sample_size) for i in sets]
ntraining = np.concatenate([i[0] for i in samples])
ntarget = np.concatenate([i[1] for i in samples])
ntest = np.concatenate([i[2] for i in sets])
ntest_target = np.concatenate([i[3] for i in sets])
ntraining, ntarget = shuffle_together(ntraining, ntarget)
ntest, ntest_target = shuffle_together(ntest, ntest_target)
return ntraining, ntarget, ntest, ntest_target
def generate_training_set(files):
targets = [h5_to_target(i) for i in files]
splits = [make_test_split(*i) for i in targets]
training, training_target, test, test_target = combine_sets(splits, sample_size = 5000)
return training, training_target, test, test_target