-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdnn2.py
326 lines (277 loc) · 14.1 KB
/
dnn2.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
# ssh gpu-0-X
# conda activate py36
import os
import sys
import time
#os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
import uproot
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.stats import norm
from scipy.optimize import curve_fit
from scipy import asarray as ar,exp
import tensorflow as tf
from tensorflow.python.eager import backprop
import pickle
from keras.callbacks import EarlyStopping, ModelCheckpoint
from keras.models import load_model
from utils.plots import *
from sklearn.metrics import confusion_matrix
from sklearn.model_selection import train_test_split
from sklearn.metrics import roc_curve, roc_auc_score
from matplotlib.backends.backend_pdf import PdfPages
from utils.var_functions import *
###################################################
# I/O #
###################################################
indir = "./samples2/"; PRE = "FULL_1204_14TeV"
outdir = "./DNN_result/" + PRE + "/LetsFind_tthh/bCat_higgs5_2Mat/" # MODIFY #
os.makedirs(outdir, exist_ok=True)
process_names = ["tthh", "tthbb", "ttbb", "ttbbbb", "ttbbcc"]
dnn1vars = [
"bJet_size",
"bJet1_pt", "bJet1_eta", "bJet1_phi", "bJet1_mass",
"bJet2_pt", "bJet2_eta", "bJet2_phi", "bJet2_mass",
"bJet3_pt", "bJet3_eta", "bJet3_phi", "bJet3_mass",
"bJet4_pt", "bJet4_eta", "bJet4_phi", "bJet4_mass",
"bJet5_pt", "bJet5_eta", "bJet5_phi", "bJet5_mass",
# bb_dr
"b1b2_dr", "b1b3_dr", "b1b4_dr", "b1b5_dr",
"b2b3_dr", "b2b4_dr", "b2b5_dr",
"b3b4_dr", "b3b5_dr",
"b4b5_dr",
]
addvars = [
# Lepton
"Lep_size",
"Lep1_pt", "Lep1_eta", "Lep1_phi", "Lep1_t",
"Lep2_pt", "Lep2_eta", "Lep2_phi", "Lep2_t",
"MET_E", # why decrease..
# Defined Kinematic vars
"bb_avg_dr", "bb_max_dr", "bb_min_dr", "b_ht", "bb_dEta_WhenMaxdR", "b_cent", "bb_max_deta", "bb_max_mass", "bb_twist",
"close_Higgs_pt", "close_Higgs_eta", "close_Higgs_phi", "close_Higgs_mass"
]
inputvars = dnn1vars + addvars
newvars = ["pred_bcat", "higgs_mass", "X_higgs"]
inputvars_2 = inputvars + newvars
openvars = inputvars
###################################################
# PreProcessing_1 #
###################################################
df_tthh = uproot.open(indir+PRE+"_tthh_di.root")["Delphes"].arrays(openvars,library="pd")
df_tthbb = uproot.open(indir+PRE+"_tthbb_di.root")["Delphes"].arrays(openvars,library="pd")
df_ttbbbb = uproot.open(indir+PRE+"_ttbbbb_di.root")["Delphes"].arrays(openvars,library="pd")
df_ttbb = uproot.open(indir+PRE+"_ttbb_di.root")["Delphes"].arrays(openvars,library="pd")
df_ttbbcc = uproot.open(indir+PRE+"_ttbbcc_di.root")["Delphes"].arrays(openvars,library="pd")
df_tthh["category"] = 0
df_tthbb["category"] = 1
df_ttbb["category"] = 2
df_ttbbbb["category"] = 3
df_ttbbcc["category"] = 4
print("Columns", df_tthh.columns)
ntthh = len(df_tthh)
ntthbb = len(df_tthbb)
nttbb = len(df_ttbb)
nttbbbb = len(df_ttbbbb)
nttbbcc = len(df_ttbbcc)
ntrain = min(ntthh, ntthbb, nttbb, nttbbbb, nttbbcc)
df_tthh = df_tthh.sample(n=ntrain).reset_index(drop=True)
df_tthbb = df_tthbb.sample(n=ntrain).reset_index(drop=True)
df_ttbb = df_ttbb.sample(n=ntrain).reset_index(drop=True)
df_ttbbbb = df_ttbbbb.sample(n=ntrain).reset_index(drop=True)
df_ttbbcc = df_ttbbcc.sample(n=ntrain).reset_index(drop=True)
# X-Y Partition
df_total = pd.concat([df_tthh, df_tthbb, df_ttbb, df_ttbbbb, df_ttbbcc])
df_total = df_total.sample(frac=1).reset_index(drop=True)
x_bCat = np.array(df_total.filter(items = dnn1vars))
x_total = np.array(df_total.filter(items = inputvars))
y_total = np.array(df_total.filter(items = ["category"]))
###################################################
# bJet Classification #
###################################################
model_dir = "DNN_result/B_1204_14TeV/bJetCassification/bCat_higgs5_2Mat/best_model.h5" ## MODIFY!!! ##
bft_model = tf.keras.models.load_model(model_dir)
bft_model.summary()
print("x_total: ", x_total); print("x_total_shape: ", x_total.shape)
pred_bcat = bft_model.predict(x_bCat); print("pred_bcat : ", pred_bcat)
pred_bcat = np.argmax(pred_bcat, axis=1) # (arg 0~10 that with the biggest prob)
### NEW VARIABLES ###
df_total["pred_bcat"] = pred_bcat
df_total["higgs_mass"] = df_total.apply(higgs_5_2, axis = 1) # MODIFY #
df_total["X_higgs"] = df_total.apply(X_higgs, axis = 1) # MODIFY #
#df_total["bfh_dr"] = df_total.apply(bfh_dr, axis = 1) # MODIFY #
### PLOTTING ###
df_plot = df_total[(df_total["category"] == 0) & (df_total["higgs_mass"]>0)] # You can draw higgs from no higgs events.
higgs_mass = np.array(df_plot.filter(items = ["higgs_mass"]))
hist, bin_edges = np.histogram(higgs_mass, bins=39, range = (0, 500), density=True)
bin_centers = (bin_edges[:-1] + bin_edges[1:]) / 2
def gaussian(x, mu, sigma, A):
return A * np.exp(-(x - mu)**2 / (2 * sigma**2))
initial_guess = [125, 30, 1] # mu, sigma, A
popt, _ = curve_fit(gaussian, bin_centers, hist, p0=initial_guess)
mu, sigma, A = popt
lower_bound = mu - sigma
upper_bound = mu + sigma
# Assuming df_new is your DataFrame containing 'close_Higgs_mass'
close_higgs_mass = np.array(df_plot.filter(items=["close_Higgs_mass"]))
hist_close, bin_edges_close = np.histogram(close_higgs_mass, bins=39, range=(0, 500), density=True)
bin_centers_close = (bin_edges_close[:-1] + bin_edges_close[1:]) / 2
# Define Gaussian function
def gaussian(x, mu, sigma, A):
return A * np.exp(-(x - mu)**2 / (2 * sigma**2))
# Initial guess for curve fitting
initial_guess = [125, 30, 1] # mu, sigma, A
# Perform curve fitting
popt_close, _ = curve_fit(gaussian, bin_centers_close, hist_close, p0=initial_guess)
# Extract parameters from the fitting
mu_close, sigma_close, A_close = popt_close
lower_bound_close = mu_close - sigma_close
upper_bound_close = mu_close + sigma_close
# Plotting
plt.hist(higgs_mass, bins=39, range=(10, 400), density=True, alpha=0.2, color='g', label='Higgs_mass Histogram')
plt.plot(bin_centers, gaussian(bin_centers, *popt), 'r-', label='Higgs_mass Fit')
plt.hist(close_higgs_mass, bins=39, range=(10, 400), density=True, alpha=0.2, color='b', label='close_Higgs_mass Histogram')
plt.plot(bin_centers_close, gaussian(bin_centers_close, *popt_close), 'b-', label='close_Higgs_mass Fit')
#plt.axvline(x=lower_bound, color='orange', linestyle='--', label='Higgs_mass 1 Sigma')
#plt.axvline(x=upper_bound, color='orange', linestyle='--')
#plt.axvline(x=lower_bound_close, color='cyan', linestyle='--', label='close_Higgs_mass 1 Sigma')
#plt.axvline(x=upper_bound_close, color='cyan', linestyle='--')
plt.axvline(x=125, color='magenta', linestyle='--', label='x=125')
plt.fill_betweenx([0, A], lower_bound, upper_bound, color='orange', alpha=0.3, label='Higgs_mass 1 Sigma Range')
plt.fill_betweenx([0, A_close], lower_bound_close, upper_bound_close, color='cyan', alpha=0.3, label='close_Higgs_mass 1 Sigma Range')
plt.xlabel('Higgs Mass [GeV]')
plt.ylabel('Probability Density')
plt.title(f'Higgs_mass: mu = {popt[0]:.2f}, sigma = {popt[1]:.2f}\n close_Higgs_mass: mu = {popt_close[0]:.2f}, sigma = {popt_close[1]:.2f}')
plt.legend()
plt.savefig(outdir+'/higgs_mass_and_close_higgs_mass_dnn.pdf')
plt.show()
'''
plt.hist(higgs_mass, bins=39, range = (10,400), density=True, alpha=0.2, color='g', label='Histogram')
plt.plot(bin_centers, gaussian(bin_centers, *popt), 'r-', label='Fit')
plt.axvline(x=lower_bound, color='b', linestyle='--', label='1 Sigma')
plt.axvline(x=upper_bound, color='b', linestyle='--')
plt.axvline(x=125, color='magenta', linestyle='--', label='x=125')
plt.fill_betweenx([0, A], lower_bound, upper_bound, color='blue', alpha=0.3, label='1 Sigma Range')
plt.xlabel('Higgs Mass [GeV]')
plt.ylabel('Probability Density')
plt.title(f'Higgs_mass: mu = {popt[0]:.2f}, sigma = {popt[1]:.2f}')
plt.legend()
plt.savefig(outdir+'/higgs_mass_dnn.pdf')
plt.show()
'''
###################################################
# PreProcessing_2 #
###################################################
# Case1 : One Hot Encoding for pred_bcat #
_x_total = df_total.filter(items = inputvars_2)
_x_ohe = pd.get_dummies(_x_total['pred_bcat'], prefix='pred_bcat', drop_first=True)
_x_total = _x_total.drop('pred_bcat', axis=1)
_x_total = pd.concat([_x_total, _x_ohe], axis = 1)
x_total = np.array(_x_total)
y_total = np.array(df_total.filter(items = ["category"]))
# Case1 : One Hot Encoding for pred_bcat #
'''
# Case 2 : No OHE, No pred_bcat #
_x_total = df_total.filter(items = inputvars_2)
_x_total = _x_total.drop('pred_bcat', axis=1)
x_total = np.array(_x_total)
y_total = np.array(df_total.filter(items = ["category"]))
# Case 2 : No OHE, No pred_bcat #
'''
print("Final x = ", x_total)
print("Final y = ", y_total)
# Data Set Partioning
ntotal = len(y_total)
train_len = int(0.7*ntotal)
x_train, x_val, y_train, y_val = train_test_split(x_total, y_total, test_size=0.3)
###################################################
# Correlation Matrix #
###################################################
# (Optional) Sample Features
#print("Plotting corr_matrix total"); plot_corrMatrix(df_total, outdir,"total")
#print("Plotting corr_matrix tthh"); plot_corrMatrix(df_tthh, outdir,"tthh")
#print("Plotting corr_matrix tthbb"); plot_corrMatrix(df_tthbb, outdir,"tthbb")
#print("Plotting corr_matrix ttbb"); plot_corrMatrix(df_ttbb, outdir,"ttbb")
#print("Plotting corr_matrix ttbbbb"); plot_corrMatrix(df_ttbbbb, outdir,"ttbbbb")
#print("Plotting corr_matrix ttbbcc"); plot_corrMatrix(df_ttbbcc, outdir,"ttbbcc")
###################################################
# Model #
###################################################
epochs = 1000; patience_epoch = 10; batch_size = 1024; print("batch size :", batch_size)
activation_function='relu'
weight_initializer = 'random_normal'
es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=patience_epoch)
mc = ModelCheckpoint(outdir+'/best_model.h5', monitor='val_loss', mode='min', save_best_only=True)
model = tf.keras.models.Sequential()
############### Input Layer ###############
model.add(tf.keras.layers.Flatten(input_shape = (x_train.shape[1],)))
############### Hidden Layer ###############
model.add(tf.keras.layers.BatchNormalization())
model.add(tf.keras.layers.Dense(50, activation=activation_function))
model.add(tf.keras.layers.BatchNormalization())
model.add(tf.keras.layers.Dense(50, activation=activation_function, kernel_regularizer='l2', kernel_initializer=weight_initializer))
############### Output Layer ###############
model.add(tf.keras.layers.Dense(len(process_names), activation="softmax"))
###################################################
############### Compile Model ###############
model.compile(optimizer=tf.keras.optimizers.Adam(clipvalue=0.5),
loss="sparse_categorical_crossentropy",
metrics = ["accuracy", "sparse_categorical_accuracy"])
model.summary()
start_time = time.time()
hist = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs,
validation_data=(x_val,y_val), callbacks=[es, mc])
end_time = time.time()
###################################################
# Prediction #
###################################################
print("# PREDICTION #")
pred_train = model.predict(x_train); print(pred_train); pred_train_arg = np.argmax(pred_train, axis=1)
pred_val = model.predict(x_val); print(pred_val); pred_val_arg = np.argmax(pred_val, axis=1)
print("Is it similar?")
print("Prediction for validation set: ", pred_val_arg)
print("Answer for train set: ", y_val.T)
train_result = pd.DataFrame(np.array([y_train.T[0], pred_train.T[0]]).T, columns=["True", "Pred"]) # True0~4,Pred0~1<"tthh"
val_result = pd.DataFrame(np.array([y_val.T[0], pred_val.T[0]]).T, columns=["True", "Pred"])
###################################################
# Confusion Matrix #
###################################################
print("# CONFUSION MATRIX #")
plot_confusion_matrix(y_val, pred_val_arg, classes=process_names,
title='Confusion matrix, without normalization', savename=outdir+"/confusion_matrix_val.pdf")
plot_confusion_matrix(y_val, pred_val_arg, classes=process_names, normalize=True,
title='Normalized confusion matrix', savename=outdir+"/norm_confusion_matrix_val.pdf")
plot_confusion_matrix(y_train, pred_train_arg, classes=process_names,
title='Confusion matrix, without normalization', savename=outdir+"/confusion_matrix_train.pdf")
plot_confusion_matrix(y_train, pred_train_arg, classes=process_names, normalize=True,
title='Normalized confusion matrix', savename=outdir+"/norm_confusion_matrix_train.pdf")
plot_output_dist(train_result, val_result,sig="tthh", savedir=outdir)
plot_performance(hist=hist, savedir=outdir)
###################################################
# Accuracy #
###################################################
print("# ACCURACY #")
train_results = model.evaluate(x_train, y_train) # Cause you set two : "accuracy", "sparse_categorical_accuracy"
train_loss = train_results[0]
train_acc = train_results[1]
print(f"Train accuracy: {train_acc * 100:.2f}%")
test_results = model.evaluate(x_val, y_val)
test_loss = test_results[0]
test_acc = test_results[1]
print(f"Test accuracy: {test_acc * 100:.2f}%")
###################################################
# Feature Importance #
###################################################
print("# FEATURE IMPORTANCE #")
model_dir = outdir + '/best_model.h5'
plot_feature_importance(model_dir, x_val, _x_total.columns, outdir)
###################################################
# Time #
###################################################
print("Number of full data: ", ntrain*5)
colnames = _x_total.columns; print("Columns :",colnames)
execution_time = end_time - start_time
print(f"execution time: {execution_time} second")
print("---Done---")