In [1]:
import pyarrow.parquet as pq
import numpy as np
import pandas as pd
from tensorflow.keras.layers import *
from tensorflow.keras.layers import Concatenate
from tensorflow.keras.models import Sequential, Model
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.utils import to_categorical
from tensorflow.keras.metrics import AUC
import tensorflow as tf
import warnings
warnings.filterwarnings('ignore')
In [2]:
file = 'QCDToGGQQ_IMGjet_RH1all_jet0_run0_n36272.test.snappy.parquet'
In [3]:
data = pq.read_table(file)
In [4]:
data = data.to_pandas()
In [5]:
data.head()
Out[5]:
| X_jets | pt | m0 | y | |
|---|---|---|---|---|
| 0 | [[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0... | 112.411095 | 21.098248 | 0.0 |
| 1 | [[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0... | 95.220406 | 14.030600 | 1.0 |
| 2 | [[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0... | 97.007317 | 17.728968 | 1.0 |
| 3 | [[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0... | 82.490311 | 14.702741 | 0.0 |
| 4 | [[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.328483... | 102.539238 | 19.456257 | 0.0 |
In [6]:
data['pt'].min()
Out[6]:
70.3982162475586
In [7]:
pt = data['pt'].to_numpy()
m0 = data['m0'].to_numpy()
y = data['y'].to_numpy()
In [8]:
X_jets = data['X_jets'].to_numpy()
In [9]:
X = np.zeros((36272,3,125,125))
In [10]:
for i in range(0,36272):
for j in range(0,3):
for k in range(0,125):
for l in range(0,125):
X[i][j][k][l] = X_jets[i][j][k][l]
In [11]:
del data
del X_jets
In [12]:
y = to_categorical(y, num_classes=2)
In [13]:
pt.shape
Out[13]:
(36272,)
In [21]:
input1 = Input(shape=(3,125,125))
input2 = Input(shape=(1,))
input3 = Input(shape=(1,))
x1 = Conv2D(3, (5,5), activation='relu', input_shape=(3, 125, 125), data_format='channels_first', padding='same')(input1)
x1 = MaxPooling2D((5,5), data_format='channels_first')(x1)
x1 = Flatten()(x1)
x1 = Dense(12, activation='relu')(x1)
x2 = Dense(4, activation='relu')(input2)
x3 = Dense(4, activation='relu')(input3)
x = Concatenate(axis=1)([x1, x2, x3])
output = Dense(2, activation='softmax')(x)
model = Model(inputs=[input1, input2, input3], outputs=output)
In [22]:
model.summary()
tf.keras.utils.plot_model(model)
Model: "model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) [(None, 3, 125, 125 0 []
)]
conv2d (Conv2D) (None, 3, 125, 125) 228 ['input_1[0][0]']
max_pooling2d (MaxPooling2D) (None, 3, 25, 25) 0 ['conv2d[0][0]']
flatten (Flatten) (None, 1875) 0 ['max_pooling2d[0][0]']
input_2 (InputLayer) [(None, 1)] 0 []
input_3 (InputLayer) [(None, 1)] 0 []
dense (Dense) (None, 12) 22512 ['flatten[0][0]']
dense_1 (Dense) (None, 4) 8 ['input_2[0][0]']
dense_2 (Dense) (None, 4) 8 ['input_3[0][0]']
concatenate (Concatenate) (None, 20) 0 ['dense[0][0]',
'dense_1[0][0]',
'dense_2[0][0]']
dense_3 (Dense) (None, 2) 42 ['concatenate[0][0]']
==================================================================================================
Total params: 22,798
Trainable params: 22,798
Non-trainable params: 0
__________________________________________________________________________________________________
Out[22]:
In [23]:
model.compile(loss="categorical_crossentropy", optimizer=Adam(learning_rate=0.0005), metrics=[AUC()])
In [24]:
with tf.device('/gpu:0'):
model.fit([X, pt, m0], y, epochs=10, batch_size=32, validation_split=0.1)
Epoch 1/10 1021/1021 [==============================] - 14s 14ms/step - loss: 6.1948 - auc: 0.5858 - val_loss: 2.0117 - val_auc: 0.6429 Epoch 2/10 1021/1021 [==============================] - 12s 12ms/step - loss: 1.3758 - auc: 0.6577 - val_loss: 0.8610 - val_auc: 0.6683 Epoch 3/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.7029 - auc: 0.7002 - val_loss: 0.6296 - val_auc: 0.7243 Epoch 4/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.6042 - auc: 0.7456 - val_loss: 0.6179 - val_auc: 0.7396 Epoch 5/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.5899 - auc: 0.7592 - val_loss: 0.5967 - val_auc: 0.7645 Epoch 6/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.5832 - auc: 0.7657 - val_loss: 0.5872 - val_auc: 0.7658 Epoch 7/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.5770 - auc: 0.7714 - val_loss: 0.5954 - val_auc: 0.7581 Epoch 8/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.5753 - auc: 0.7734 - val_loss: 0.5790 - val_auc: 0.7739 Epoch 9/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.5692 - auc: 0.7786 - val_loss: 0.5873 - val_auc: 0.7703 Epoch 10/10 1021/1021 [==============================] - 12s 12ms/step - loss: 0.5630 - auc: 0.7851 - val_loss: 0.5882 - val_auc: 0.7710
Train AUC Score: 0.785
Validation AUC Score: 0.771