notebook.community

Edit and run

This program will not generate the jet images, it will only train the autoencoder and evaluate the results. The jet images can be found in:

https://drive.google.com/drive/folders/1i5DY9duzDuumQz636u5YQeYQEt_7TYa8?usp=sharing

Please download those images to your google drive and use the colab - drive integration.

A program to generate jet images is available at

https://github.com/aravindhv10/CPP_Wrappers/blob/master/AntiQCD4/JetImageFormation.hh

in the form of the class BoxImageGen. The images used in this program were produced using BoxImageGen<40,float,true> with the ratio $m_J/E_J=0.5$.



In [0]:

    
# This program will not generate the jet images, it will only train the autoencoder
# and evaluate the results. The jet images can be found in:
# https://drive.google.com/drive/folders/1i5DY9duzDuumQz636u5YQeYQEt_7TYa8?usp=sharing
# Please download those images to your google drive and use the colab - drive integration.
import lzma
from google.colab import drive
import numpy as np
import tensorflow as tf
import keras
from keras import backend as K
from keras.layers import Input, Dense
from keras.models import Model
import matplotlib.pyplot as plt

def READ_XZ (filename):
    file = lzma.LZMAFile(filename)
    type_bytes = file.read(-1)
    type_array = np.frombuffer(type_bytes,dtype='float32')                                                
    return type_array

def Count(array,val):
  count = 0.0
  for e in range(array.shape[0]):
    if array[e]>val :
      count=count+1.0
  return count / array.shape[0]

width=40
batch_size=200
ModelName = "Model_40_24_8_24_40_40"

config = tf.ConfigProto( device_count = {'GPU': 1 , 'CPU': 2} ) 
sess = tf.Session(config=config)
keras.backend.set_session(sess)
K.tensorflow_backend._get_available_gpus()









    





The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.

We recommend you upgrade now 
or ensure your notebook will continue to use TensorFlow 1.x via the %tensorflow_version 1.x magic:
.







    



Using TensorFlow backend.






    



WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:190: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:207: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.







    Out[0]:





['/job:localhost/replica:0/task:0/device:GPU:0']

Defining network architecture (we use `Arch-2`)

We also define some functions to make training convinent here.



In [0]:

    
# this is our input placeholder
input_img = Input(shape=(width*width,))

# "encoded" is the encoded representation of the input
Layer1 = Dense(24*24, activation='relu')(input_img)
Layer2 = Dense(8*8, activation='relu')(Layer1)
Layer3 = Dense(24*24, activation='relu')(Layer2)
Layer4 = Dense(40*40, activation='relu')(Layer3)
Out = Dense(40*40, activation='softmax')(Layer4)

# this model maps an input to its reconstruction
autoencoder = Model(input_img, Out)
autoencoder.compile(optimizer='adam', loss='mean_squared_error')

def EvalOnFile (InFileName,OutFileName):
  data = READ_XZ (InFileName)
  x_train = data.reshape(-1,width*width)
  x_out = autoencoder.predict(x_train,200,use_multiprocessing=True)
  diff = x_train - x_out
  lrnorm = np.ones((diff.shape[0]))
  for e in range(diff.shape[0]):
    lrnorm[e] = np.linalg.norm(diff[e])
  lrnorm.tofile(OutFileName)
  print(lrnorm.shape)

def TrainOnFile (filename,testfilename,totalepochs):
  data = READ_XZ (filename)
  x_train = data.reshape(-1,width*width)
  datatest = READ_XZ (testfilename)
  x_test = datatest.reshape(-1,width*width)
  autoencoder.fit(
      x_train, x_train, epochs=totalepochs,
      batch_size=200, shuffle=True,
      validation_data=(x_test, x_test)
  )
  autoencoder.save(ModelName)









    



WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:793: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.

Mounting folder from Google Drive:



In [0]:

    
# Please download the files from the link below and appropriately change this program:
# https://drive.google.com/drive/folders/1i5DY9duzDuumQz636u5YQeYQEt_7TYa8?usp=sharing
drive.mount('/gdrive')
%cd /gdrive









    



Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly

Enter your authorization code:
··········
Mounted at /gdrive
/gdrive

Verify the files are correctly mounted and available:



In [0]:

    
%cd /gdrive/My\ Drive/JetImages/QCD/
!ls ./TEST/BoxImages/0.xz
!ls ./TRAIN/BoxImages/0.xz









    



/gdrive/My Drive/JetImages/QCD
./TEST/BoxImages/0.xz
./TRAIN/BoxImages/0.xz

Load the model in case a trained one is already available:



In [0]:

    
%cd /gdrive/My Drive/JetImages/QCD
autoencoder = keras.models.load_model(ModelName)









    



/gdrive/My Drive/JetImages/QCD
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:216: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:223: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1033: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1020: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.

The training step:



In [0]:

    
%cd /gdrive/My Drive/JetImages/QCD
for e in range(4):
  TrainOnFile("./TRAIN/BoxImages/0.xz","./TEST/BoxImages/0.xz",10)
  TrainOnFile("./TRAIN/BoxImages/1.xz","./TEST/BoxImages/1.xz",10)
  TrainOnFile("./TRAIN/BoxImages/2.xz","./TEST/BoxImages/2.xz",10)
  TrainOnFile("./TRAIN/BoxImages/3.xz","./TEST/BoxImages/3.xz",10)
  TrainOnFile("./TRAIN/BoxImages/4.xz","./TEST/BoxImages/4.xz",10)
  TrainOnFile("./TRAIN/BoxImages/5.xz","./TEST/BoxImages/5.xz",10)
  TrainOnFile("./TRAIN/BoxImages/6.xz","./TEST/BoxImages/6.xz",10)
  TrainOnFile("./TRAIN/BoxImages/7.xz","./TEST/BoxImages/7.xz",10)
  TrainOnFile("./TRAIN/BoxImages/8.xz","./TEST/BoxImages/8.xz",10)
  TrainOnFile("./TRAIN/BoxImages/9.xz","./TEST/BoxImages/9.xz",10)
  TrainOnFile("./TRAIN/BoxImages/10.xz","./TEST/BoxImages/10.xz",10)
  TrainOnFile("./TRAIN/BoxImages/11.xz","./TEST/BoxImages/11.xz",10)
  TrainOnFile("./TRAIN/BoxImages/12.xz","./TEST/BoxImages/12.xz",10)
  TrainOnFile("./TRAIN/BoxImages/13.xz","./TEST/BoxImages/13.xz",10)
  TrainOnFile("./TRAIN/BoxImages/14.xz","./TEST/BoxImages/14.xz",10)
  TrainOnFile("./TRAIN/BoxImages/15.xz","./TEST/BoxImages/15.xz",10)









    



WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1033: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1020: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.

Train on 100000 samples, validate on 20000 samples
Epoch 1/10
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:216: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:223: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.

100000/100000 [==============================] - 5s 47us/step - loss: 1.0627e-04 - val_loss: 1.0634e-04
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.0621e-04 - val_loss: 1.0624e-04
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.6311e-05 - val_loss: 5.9276e-05
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8847e-05 - val_loss: 3.0651e-05
Epoch 5/10
100000/100000 [==============================] - 4s 35us/step - loss: 2.6664e-05 - val_loss: 2.3519e-05
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 2.1724e-05 - val_loss: 2.0635e-05
Epoch 7/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.9779e-05 - val_loss: 1.9115e-05
Epoch 8/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.8726e-05 - val_loss: 1.8336e-05
Epoch 9/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.7859e-05 - val_loss: 1.7425e-05
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.6875e-05 - val_loss: 1.6549e-05
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.6198e-05 - val_loss: 1.6035e-05
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.5692e-05 - val_loss: 1.5579e-05
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.5201e-05 - val_loss: 1.5058e-05
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.4754e-05 - val_loss: 1.4708e-05
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.4372e-05 - val_loss: 1.4255e-05
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.4005e-05 - val_loss: 1.3937e-05
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.3614e-05 - val_loss: 1.3538e-05
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.3265e-05 - val_loss: 1.3252e-05
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.2968e-05 - val_loss: 1.2967e-05
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.2692e-05 - val_loss: 1.2683e-05
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.2533e-05 - val_loss: 1.2338e-05
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.2328e-05 - val_loss: 1.2162e-05
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.2175e-05 - val_loss: 1.2048e-05
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.2025e-05 - val_loss: 1.1900e-05
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1879e-05 - val_loss: 1.1748e-05
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1742e-05 - val_loss: 1.1652e-05
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1599e-05 - val_loss: 1.1510e-05
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1478e-05 - val_loss: 1.1388e-05
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1355e-05 - val_loss: 1.1277e-05
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1220e-05 - val_loss: 1.1140e-05
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.1077e-05 - val_loss: 1.1106e-05
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.0968e-05 - val_loss: 1.1058e-05
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.0872e-05 - val_loss: 1.0956e-05
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.0769e-05 - val_loss: 1.0822e-05
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.0672e-05 - val_loss: 1.0728e-05
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.0576e-05 - val_loss: 1.0622e-05
Epoch 7/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.0481e-05 - val_loss: 1.0561e-05
Epoch 8/10
100000/100000 [==============================] - 4s 35us/step - loss: 1.0400e-05 - val_loss: 1.0435e-05
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.0313e-05 - val_loss: 1.0432e-05
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.0241e-05 - val_loss: 1.0309e-05
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.0301e-05 - val_loss: 1.0329e-05
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 1.0206e-05 - val_loss: 1.0284e-05
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.0118e-05 - val_loss: 1.0169e-05
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 1.0041e-05 - val_loss: 1.0099e-05
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.9610e-06 - val_loss: 1.0036e-05
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.8882e-06 - val_loss: 9.9848e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.8160e-06 - val_loss: 9.9071e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.7435e-06 - val_loss: 9.8389e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.6901e-06 - val_loss: 9.8035e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.6151e-06 - val_loss: 9.7470e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 9.6597e-06 - val_loss: 9.5944e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.5865e-06 - val_loss: 9.5060e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.5218e-06 - val_loss: 9.4821e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.4676e-06 - val_loss: 9.4151e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.4058e-06 - val_loss: 9.3492e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.3517e-06 - val_loss: 9.3045e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.2953e-06 - val_loss: 9.2750e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 9.2446e-06 - val_loss: 9.2007e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.1910e-06 - val_loss: 9.1551e-06
Epoch 10/10
100000/100000 [==============================] - 4s 36us/step - loss: 9.1326e-06 - val_loss: 9.0958e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 9.0694e-06 - val_loss: 8.9412e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 8.9925e-06 - val_loss: 8.9293e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.9263e-06 - val_loss: 8.8557e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.8674e-06 - val_loss: 8.7999e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.8045e-06 - val_loss: 8.7643e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.7496e-06 - val_loss: 8.7158e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.6954e-06 - val_loss: 8.6511e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.6416e-06 - val_loss: 8.5872e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.5878e-06 - val_loss: 8.5389e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.5375e-06 - val_loss: 8.5051e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.6332e-06 - val_loss: 8.4847e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 8.5704e-06 - val_loss: 8.4284e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.5181e-06 - val_loss: 8.3927e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.4735e-06 - val_loss: 8.3690e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.4277e-06 - val_loss: 8.3327e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.3883e-06 - val_loss: 8.2933e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.3515e-06 - val_loss: 8.2765e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.3095e-06 - val_loss: 8.2232e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.2703e-06 - val_loss: 8.2039e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.2308e-06 - val_loss: 8.1521e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 8.2694e-06 - val_loss: 8.1410e-06
Epoch 2/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.2127e-06 - val_loss: 8.1180e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.1660e-06 - val_loss: 8.0654e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.1326e-06 - val_loss: 8.0324e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 8.0984e-06 - val_loss: 8.0000e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.0552e-06 - val_loss: 7.9752e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 8.0204e-06 - val_loss: 7.9470e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.9909e-06 - val_loss: 7.9415e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.9507e-06 - val_loss: 7.9015e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.9179e-06 - val_loss: 7.8523e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.8914e-06 - val_loss: 7.9466e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 7.8393e-06 - val_loss: 7.8881e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.7976e-06 - val_loss: 7.8684e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.7627e-06 - val_loss: 7.8503e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.7275e-06 - val_loss: 7.8135e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.6963e-06 - val_loss: 7.8002e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.6688e-06 - val_loss: 7.7715e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.6344e-06 - val_loss: 7.7428e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.6070e-06 - val_loss: 7.7325e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.5813e-06 - val_loss: 7.6769e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.6974e-06 - val_loss: 7.7758e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.6463e-06 - val_loss: 7.7287e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.6042e-06 - val_loss: 7.6932e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.5600e-06 - val_loss: 7.6740e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.5252e-06 - val_loss: 7.6426e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.4878e-06 - val_loss: 7.5978e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.4533e-06 - val_loss: 7.5745e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.4188e-06 - val_loss: 7.5509e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.3882e-06 - val_loss: 7.5314e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.3653e-06 - val_loss: 7.5107e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 7.4654e-06 - val_loss: 7.4843e-06
Epoch 2/10
100000/100000 [==============================] - 4s 36us/step - loss: 7.4093e-06 - val_loss: 7.4612e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.3711e-06 - val_loss: 7.4096e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.3407e-06 - val_loss: 7.3868e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.3092e-06 - val_loss: 7.3954e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.2771e-06 - val_loss: 7.3407e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.2471e-06 - val_loss: 7.3308e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.2182e-06 - val_loss: 7.3391e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.1925e-06 - val_loss: 7.2741e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 7.1627e-06 - val_loss: 7.2633e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.2044e-06 - val_loss: 7.1965e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 7.1578e-06 - val_loss: 7.1655e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.1188e-06 - val_loss: 7.1600e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.0860e-06 - val_loss: 7.1196e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.0519e-06 - val_loss: 7.0941e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 7.0244e-06 - val_loss: 7.0970e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.9929e-06 - val_loss: 7.0548e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.9656e-06 - val_loss: 7.0330e-06
Epoch 9/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.9422e-06 - val_loss: 7.0430e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.9159e-06 - val_loss: 6.9979e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 7.0328e-06 - val_loss: 7.0887e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.9908e-06 - val_loss: 7.0634e-06
Epoch 3/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.9505e-06 - val_loss: 7.0181e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.9235e-06 - val_loss: 7.0103e-06
Epoch 5/10
100000/100000 [==============================] - 4s 36us/step - loss: 6.8925e-06 - val_loss: 6.9893e-06
Epoch 6/10
100000/100000 [==============================] - 4s 36us/step - loss: 6.8637e-06 - val_loss: 6.9663e-06
Epoch 7/10
100000/100000 [==============================] - 4s 36us/step - loss: 6.8394e-06 - val_loss: 6.9482e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.8126e-06 - val_loss: 6.9441e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.7858e-06 - val_loss: 6.9089e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.7673e-06 - val_loss: 6.9239e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.8281e-06 - val_loss: 6.9183e-06
Epoch 2/10
100000/100000 [==============================] - 4s 36us/step - loss: 6.7757e-06 - val_loss: 6.9197e-06
Epoch 3/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.7440e-06 - val_loss: 6.8796e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.7104e-06 - val_loss: 6.8793e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.6781e-06 - val_loss: 6.8427e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.6552e-06 - val_loss: 6.8433e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.6260e-06 - val_loss: 6.8170e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.6042e-06 - val_loss: 6.7831e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.5781e-06 - val_loss: 6.7922e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.5556e-06 - val_loss: 6.7610e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.6825e-06 - val_loss: 6.6412e-06
Epoch 2/10
100000/100000 [==============================] - 4s 36us/step - loss: 6.6337e-06 - val_loss: 6.6140e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.5968e-06 - val_loss: 6.5681e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.5670e-06 - val_loss: 6.5782e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.5410e-06 - val_loss: 6.5517e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.5158e-06 - val_loss: 6.5239e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.4899e-06 - val_loss: 6.5228e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.4716e-06 - val_loss: 6.4835e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.4429e-06 - val_loss: 6.4770e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.4198e-06 - val_loss: 6.4579e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.4815e-06 - val_loss: 6.5513e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.4325e-06 - val_loss: 6.5042e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.4013e-06 - val_loss: 6.4848e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.3749e-06 - val_loss: 6.4675e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.3466e-06 - val_loss: 6.4764e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.3193e-06 - val_loss: 6.4324e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2993e-06 - val_loss: 6.4277e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2789e-06 - val_loss: 6.4256e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2566e-06 - val_loss: 6.3951e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2318e-06 - val_loss: 6.3767e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.3536e-06 - val_loss: 6.4004e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.3058e-06 - val_loss: 6.4068e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2689e-06 - val_loss: 6.3697e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2398e-06 - val_loss: 6.3466e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2140e-06 - val_loss: 6.3381e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.1893e-06 - val_loss: 6.3315e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.1658e-06 - val_loss: 6.3137e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.1429e-06 - val_loss: 6.3160e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.1231e-06 - val_loss: 6.3029e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.0992e-06 - val_loss: 6.2793e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.2250e-06 - val_loss: 6.1548e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.1725e-06 - val_loss: 6.1214e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.1393e-06 - val_loss: 6.1088e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.1133e-06 - val_loss: 6.1069e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.0897e-06 - val_loss: 6.0755e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.0630e-06 - val_loss: 6.0765e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 6.0446e-06 - val_loss: 6.0670e-06
Epoch 8/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.0200e-06 - val_loss: 6.0548e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.0055e-06 - val_loss: 6.0430e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.9863e-06 - val_loss: 6.0416e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 6.0644e-06 - val_loss: 6.0228e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 6.0155e-06 - val_loss: 6.0241e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.9828e-06 - val_loss: 6.0255e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.9583e-06 - val_loss: 5.9870e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.9355e-06 - val_loss: 5.9774e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.9136e-06 - val_loss: 5.9729e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.8943e-06 - val_loss: 5.9601e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.8738e-06 - val_loss: 5.9456e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.8571e-06 - val_loss: 5.9446e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.8375e-06 - val_loss: 5.9362e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.9924e-06 - val_loss: 5.9992e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.9459e-06 - val_loss: 5.9655e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.9129e-06 - val_loss: 5.9611e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.8861e-06 - val_loss: 5.9516e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.8622e-06 - val_loss: 5.9254e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.8440e-06 - val_loss: 5.9098e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.8242e-06 - val_loss: 5.9160e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.8031e-06 - val_loss: 5.8952e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7874e-06 - val_loss: 5.8925e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7692e-06 - val_loss: 5.8968e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.9092e-06 - val_loss: 5.8486e-06
Epoch 2/10
100000/100000 [==============================] - 4s 36us/step - loss: 5.8597e-06 - val_loss: 5.8357e-06
Epoch 3/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.8281e-06 - val_loss: 5.8131e-06
Epoch 4/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.8037e-06 - val_loss: 5.8046e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7819e-06 - val_loss: 5.7979e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7592e-06 - val_loss: 5.7942e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7411e-06 - val_loss: 5.7807e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7228e-06 - val_loss: 5.7675e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.7048e-06 - val_loss: 5.7445e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.6852e-06 - val_loss: 5.7542e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.7370e-06 - val_loss: 5.7314e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.6865e-06 - val_loss: 5.7044e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.6542e-06 - val_loss: 5.6864e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.6306e-06 - val_loss: 5.6877e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.6065e-06 - val_loss: 5.6662e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5914e-06 - val_loss: 5.6563e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5708e-06 - val_loss: 5.6287e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5556e-06 - val_loss: 5.6382e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5392e-06 - val_loss: 5.6265e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5213e-06 - val_loss: 5.6172e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.6870e-06 - val_loss: 5.5964e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.6398e-06 - val_loss: 5.5765e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.6101e-06 - val_loss: 5.5588e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5850e-06 - val_loss: 5.5390e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5624e-06 - val_loss: 5.5561e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5431e-06 - val_loss: 5.5327e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5231e-06 - val_loss: 5.5230e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5076e-06 - val_loss: 5.5085e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.4896e-06 - val_loss: 5.5123e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.4759e-06 - val_loss: 5.5059e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.5761e-06 - val_loss: 5.5068e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.5272e-06 - val_loss: 5.4766e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.4969e-06 - val_loss: 5.4529e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.4734e-06 - val_loss: 5.4439e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.4558e-06 - val_loss: 5.4382e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.4338e-06 - val_loss: 5.4500e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.4168e-06 - val_loss: 5.4209e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3998e-06 - val_loss: 5.4077e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3837e-06 - val_loss: 5.3950e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.3681e-06 - val_loss: 5.3838e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.4388e-06 - val_loss: 5.4599e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.3900e-06 - val_loss: 5.4405e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3622e-06 - val_loss: 5.4426e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3432e-06 - val_loss: 5.4321e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3210e-06 - val_loss: 5.4222e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3065e-06 - val_loss: 5.4160e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2857e-06 - val_loss: 5.4105e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2709e-06 - val_loss: 5.4038e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2555e-06 - val_loss: 5.3875e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2384e-06 - val_loss: 5.3855e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3911e-06 - val_loss: 5.4330e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.3371e-06 - val_loss: 5.4047e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.3101e-06 - val_loss: 5.4132e-06
Epoch 4/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.2843e-06 - val_loss: 5.3669e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.2630e-06 - val_loss: 5.3717e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2441e-06 - val_loss: 5.3635e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2264e-06 - val_loss: 5.3552e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2077e-06 - val_loss: 5.3602e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1933e-06 - val_loss: 5.3482e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1762e-06 - val_loss: 5.3343e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.3363e-06 - val_loss: 5.3919e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.2843e-06 - val_loss: 5.3832e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2553e-06 - val_loss: 5.3564e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2286e-06 - val_loss: 5.3507e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2095e-06 - val_loss: 5.3453e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1904e-06 - val_loss: 5.3430e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1729e-06 - val_loss: 5.3184e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1589e-06 - val_loss: 5.3025e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1407e-06 - val_loss: 5.3049e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1283e-06 - val_loss: 5.3025e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.2155e-06 - val_loss: 5.2346e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.1641e-06 - val_loss: 5.2204e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1334e-06 - val_loss: 5.2056e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1130e-06 - val_loss: 5.1969e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.0923e-06 - val_loss: 5.1899e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0747e-06 - val_loss: 5.1769e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0592e-06 - val_loss: 5.1780e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0452e-06 - val_loss: 5.1649e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0298e-06 - val_loss: 5.1630e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0201e-06 - val_loss: 5.1639e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.1790e-06 - val_loss: 5.2306e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 5.1245e-06 - val_loss: 5.2036e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0931e-06 - val_loss: 5.1811e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0765e-06 - val_loss: 5.1883e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0555e-06 - val_loss: 5.1843e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0359e-06 - val_loss: 5.1706e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0191e-06 - val_loss: 5.1624e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0029e-06 - val_loss: 5.1413e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9876e-06 - val_loss: 5.1379e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9751e-06 - val_loss: 5.1334e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.1047e-06 - val_loss: 5.1467e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.0515e-06 - val_loss: 5.1359e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0183e-06 - val_loss: 5.1411e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9954e-06 - val_loss: 5.1151e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9775e-06 - val_loss: 5.1051e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9589e-06 - val_loss: 5.0928e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.9439e-06 - val_loss: 5.0983e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9290e-06 - val_loss: 5.0802e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9128e-06 - val_loss: 5.0760e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8980e-06 - val_loss: 5.0788e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 5.0601e-06 - val_loss: 5.0324e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 5.0087e-06 - val_loss: 5.0317e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9790e-06 - val_loss: 4.9943e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9547e-06 - val_loss: 4.9931e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.9326e-06 - val_loss: 4.9846e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9144e-06 - val_loss: 4.9776e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9033e-06 - val_loss: 4.9744e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8874e-06 - val_loss: 4.9685e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8739e-06 - val_loss: 4.9660e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8604e-06 - val_loss: 4.9702e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9566e-06 - val_loss: 4.9989e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.8983e-06 - val_loss: 4.9768e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8708e-06 - val_loss: 4.9651e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8476e-06 - val_loss: 4.9752e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8293e-06 - val_loss: 4.9561e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8117e-06 - val_loss: 4.9563e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7957e-06 - val_loss: 4.9460e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7822e-06 - val_loss: 4.9444e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7673e-06 - val_loss: 4.9465e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7569e-06 - val_loss: 4.9391e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.9263e-06 - val_loss: 4.9536e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.8737e-06 - val_loss: 4.9266e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8434e-06 - val_loss: 4.9299e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.8192e-06 - val_loss: 4.9247e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7987e-06 - val_loss: 4.9180e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7811e-06 - val_loss: 4.9137e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7659e-06 - val_loss: 4.8999e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.7488e-06 - val_loss: 4.9109e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7352e-06 - val_loss: 4.8977e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7233e-06 - val_loss: 4.8722e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.8617e-06 - val_loss: 4.8314e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.8111e-06 - val_loss: 4.8097e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7790e-06 - val_loss: 4.8009e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7539e-06 - val_loss: 4.7964e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7321e-06 - val_loss: 4.7900e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7168e-06 - val_loss: 4.7921e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7030e-06 - val_loss: 4.7867e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6890e-06 - val_loss: 4.7780e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6751e-06 - val_loss: 4.7963e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6606e-06 - val_loss: 4.7675e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7931e-06 - val_loss: 4.7816e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.7389e-06 - val_loss: 4.7665e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7118e-06 - val_loss: 4.7766e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6889e-06 - val_loss: 4.7610e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6689e-06 - val_loss: 4.7595e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6529e-06 - val_loss: 4.7452e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.6375e-06 - val_loss: 4.7451e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6236e-06 - val_loss: 4.7351e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6111e-06 - val_loss: 4.7344e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5981e-06 - val_loss: 4.7293e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7873e-06 - val_loss: 4.7585e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.7327e-06 - val_loss: 4.7247e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.7040e-06 - val_loss: 4.7173e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6777e-06 - val_loss: 4.7211e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6630e-06 - val_loss: 4.7047e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6471e-06 - val_loss: 4.7119e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6312e-06 - val_loss: 4.6956e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6167e-06 - val_loss: 4.6966e-06
Epoch 9/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.6045e-06 - val_loss: 4.6918e-06
Epoch 10/10
100000/100000 [==============================] - 4s 36us/step - loss: 4.5934e-06 - val_loss: 4.6861e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.7610e-06 - val_loss: 4.6911e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.7092e-06 - val_loss: 4.6886e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6826e-06 - val_loss: 4.6868e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6576e-06 - val_loss: 4.6642e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6394e-06 - val_loss: 4.6756e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6233e-06 - val_loss: 4.6646e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6074e-06 - val_loss: 4.6507e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5920e-06 - val_loss: 4.6505e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5776e-06 - val_loss: 4.6456e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5673e-06 - val_loss: 4.6372e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6480e-06 - val_loss: 4.6261e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.5947e-06 - val_loss: 4.6169e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5646e-06 - val_loss: 4.6006e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.5429e-06 - val_loss: 4.5866e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5247e-06 - val_loss: 4.5874e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5072e-06 - val_loss: 4.5743e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4923e-06 - val_loss: 4.5799e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4804e-06 - val_loss: 4.5707e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.4687e-06 - val_loss: 4.5654e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4552e-06 - val_loss: 4.5830e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.6377e-06 - val_loss: 4.6119e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.5849e-06 - val_loss: 4.5783e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5553e-06 - val_loss: 4.5685e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5336e-06 - val_loss: 4.5691e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5137e-06 - val_loss: 4.5550e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4951e-06 - val_loss: 4.5450e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4833e-06 - val_loss: 4.5532e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4682e-06 - val_loss: 4.5582e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.4569e-06 - val_loss: 4.5511e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4436e-06 - val_loss: 4.5417e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.5831e-06 - val_loss: 4.5298e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.5268e-06 - val_loss: 4.5041e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4952e-06 - val_loss: 4.5020e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4757e-06 - val_loss: 4.4920e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4567e-06 - val_loss: 4.4930e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4397e-06 - val_loss: 4.4745e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.4255e-06 - val_loss: 4.4805e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4139e-06 - val_loss: 4.4777e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3996e-06 - val_loss: 4.4903e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3873e-06 - val_loss: 4.4651e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4925e-06 - val_loss: 4.5415e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.4378e-06 - val_loss: 4.5214e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4082e-06 - val_loss: 4.5073e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3880e-06 - val_loss: 4.5059e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3698e-06 - val_loss: 4.4968e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3529e-06 - val_loss: 4.4908e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3402e-06 - val_loss: 4.4985e-06
Epoch 8/10
100000/100000 [==============================] - 4s 36us/step - loss: 4.3248e-06 - val_loss: 4.5085e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.3151e-06 - val_loss: 4.4907e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3019e-06 - val_loss: 4.4777e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4721e-06 - val_loss: 4.5017e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.4197e-06 - val_loss: 4.4919e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3920e-06 - val_loss: 4.4799e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3706e-06 - val_loss: 4.4774e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3529e-06 - val_loss: 4.4860e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3379e-06 - val_loss: 4.4678e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3228e-06 - val_loss: 4.4652e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3100e-06 - val_loss: 4.4683e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2986e-06 - val_loss: 4.4657e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2898e-06 - val_loss: 4.4690e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.4704e-06 - val_loss: 4.4856e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.4059e-06 - val_loss: 4.4940e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3788e-06 - val_loss: 4.4966e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.3551e-06 - val_loss: 4.4717e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3364e-06 - val_loss: 4.4811e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.3215e-06 - val_loss: 4.4723e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3068e-06 - val_loss: 4.4576e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2943e-06 - val_loss: 4.4610e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2823e-06 - val_loss: 4.4595e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2716e-06 - val_loss: 4.4655e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3898e-06 - val_loss: 4.4165e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.3355e-06 - val_loss: 4.4067e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3048e-06 - val_loss: 4.3951e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2852e-06 - val_loss: 4.3961e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2682e-06 - val_loss: 4.3659e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2526e-06 - val_loss: 4.3789e-06
Epoch 7/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.2390e-06 - val_loss: 4.3658e-06
Epoch 8/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.2282e-06 - val_loss: 4.3693e-06
Epoch 9/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.2166e-06 - val_loss: 4.3734e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2044e-06 - val_loss: 4.3598e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.3712e-06 - val_loss: 4.4112e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.3117e-06 - val_loss: 4.4072e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2852e-06 - val_loss: 4.3778e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2631e-06 - val_loss: 4.3869e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2468e-06 - val_loss: 4.3742e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2331e-06 - val_loss: 4.3816e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2189e-06 - val_loss: 4.3857e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2063e-06 - val_loss: 4.3828e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1942e-06 - val_loss: 4.3836e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1886e-06 - val_loss: 4.3743e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3337e-06 - val_loss: 4.3624e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2803e-06 - val_loss: 4.3556e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2493e-06 - val_loss: 4.3393e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2277e-06 - val_loss: 4.3479e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2135e-06 - val_loss: 4.3424e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1981e-06 - val_loss: 4.3369e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1849e-06 - val_loss: 4.3352e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1749e-06 - val_loss: 4.3269e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1631e-06 - val_loss: 4.3301e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1537e-06 - val_loss: 4.3418e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.3280e-06 - val_loss: 4.3101e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.2686e-06 - val_loss: 4.2818e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2404e-06 - val_loss: 4.2843e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2203e-06 - val_loss: 4.2797e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2011e-06 - val_loss: 4.2753e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1884e-06 - val_loss: 4.2623e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1746e-06 - val_loss: 4.2644e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1642e-06 - val_loss: 4.2751e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1537e-06 - val_loss: 4.2674e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1429e-06 - val_loss: 4.2636e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2562e-06 - val_loss: 4.3086e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.2001e-06 - val_loss: 4.2986e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1728e-06 - val_loss: 4.2787e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1533e-06 - val_loss: 4.2827e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1348e-06 - val_loss: 4.2680e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1225e-06 - val_loss: 4.2780e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1127e-06 - val_loss: 4.2728e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1001e-06 - val_loss: 4.2575e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0859e-06 - val_loss: 4.2563e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0775e-06 - val_loss: 4.2511e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.2469e-06 - val_loss: 4.3012e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1898e-06 - val_loss: 4.2815e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1616e-06 - val_loss: 4.2755e-06
Epoch 4/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.1395e-06 - val_loss: 4.2739e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1211e-06 - val_loss: 4.2637e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1088e-06 - val_loss: 4.2687e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.0964e-06 - val_loss: 4.2584e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0841e-06 - val_loss: 4.2557e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0724e-06 - val_loss: 4.2622e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0619e-06 - val_loss: 4.2548e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.2095e-06 - val_loss: 4.2038e-06
Epoch 2/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1509e-06 - val_loss: 4.1826e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1252e-06 - val_loss: 4.1706e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1033e-06 - val_loss: 4.1580e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0867e-06 - val_loss: 4.1739e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0721e-06 - val_loss: 4.1620e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0586e-06 - val_loss: 4.1723e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0463e-06 - val_loss: 4.1607e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0361e-06 - val_loss: 4.1656e-06
Epoch 10/10
100000/100000 [==============================] - 4s 36us/step - loss: 4.0253e-06 - val_loss: 4.1566e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1664e-06 - val_loss: 4.1569e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1107e-06 - val_loss: 4.1422e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0822e-06 - val_loss: 4.1363e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0613e-06 - val_loss: 4.1426e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0448e-06 - val_loss: 4.1351e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0302e-06 - val_loss: 4.1313e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0172e-06 - val_loss: 4.1331e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0049e-06 - val_loss: 4.1352e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9937e-06 - val_loss: 4.1290e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9839e-06 - val_loss: 4.1251e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1751e-06 - val_loss: 4.1386e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.1191e-06 - val_loss: 4.1264e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0886e-06 - val_loss: 4.1179e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.0705e-06 - val_loss: 4.1161e-06
Epoch 5/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.0535e-06 - val_loss: 4.1258e-06
Epoch 6/10
100000/100000 [==============================] - 4s 36us/step - loss: 4.0405e-06 - val_loss: 4.1153e-06
Epoch 7/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.0284e-06 - val_loss: 4.1079e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0150e-06 - val_loss: 4.1119e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0027e-06 - val_loss: 4.1047e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9938e-06 - val_loss: 4.1038e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.1607e-06 - val_loss: 4.1225e-06
Epoch 2/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.1051e-06 - val_loss: 4.1076e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0775e-06 - val_loss: 4.0995e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0585e-06 - val_loss: 4.1017e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0438e-06 - val_loss: 4.0982e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0267e-06 - val_loss: 4.0915e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0154e-06 - val_loss: 4.0917e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0019e-06 - val_loss: 4.0857e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9914e-06 - val_loss: 4.0843e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9838e-06 - val_loss: 4.1000e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0903e-06 - val_loss: 4.0627e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.0316e-06 - val_loss: 4.0572e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 4.0071e-06 - val_loss: 4.0372e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9855e-06 - val_loss: 4.0476e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9677e-06 - val_loss: 4.0285e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9532e-06 - val_loss: 4.0327e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9440e-06 - val_loss: 4.0225e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9312e-06 - val_loss: 4.0198e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9201e-06 - val_loss: 4.0228e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9105e-06 - val_loss: 4.0233e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0872e-06 - val_loss: 4.0456e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.0322e-06 - val_loss: 4.0401e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 4.0066e-06 - val_loss: 4.0339e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9842e-06 - val_loss: 4.0333e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9670e-06 - val_loss: 4.0292e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9543e-06 - val_loss: 4.0228e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9418e-06 - val_loss: 4.0189e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9329e-06 - val_loss: 4.0293e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9222e-06 - val_loss: 4.0222e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9111e-06 - val_loss: 4.0239e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 4.0538e-06 - val_loss: 4.0234e-06
Epoch 2/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9980e-06 - val_loss: 4.0009e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9700e-06 - val_loss: 3.9978e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9465e-06 - val_loss: 4.0044e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9333e-06 - val_loss: 3.9883e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9195e-06 - val_loss: 3.9850e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9058e-06 - val_loss: 3.9896e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8944e-06 - val_loss: 3.9849e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8824e-06 - val_loss: 3.9828e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8755e-06 - val_loss: 3.9778e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.9886e-06 - val_loss: 4.0356e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.9339e-06 - val_loss: 4.0115e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9049e-06 - val_loss: 4.0100e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8870e-06 - val_loss: 4.0144e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8692e-06 - val_loss: 4.0069e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8541e-06 - val_loss: 4.0192e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8436e-06 - val_loss: 3.9928e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8328e-06 - val_loss: 3.9950e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8215e-06 - val_loss: 4.0076e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8134e-06 - val_loss: 4.0000e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9845e-06 - val_loss: 4.0305e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9261e-06 - val_loss: 4.0243e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9012e-06 - val_loss: 4.0065e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8783e-06 - val_loss: 4.0130e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8642e-06 - val_loss: 3.9984e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8503e-06 - val_loss: 3.9904e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8380e-06 - val_loss: 3.9996e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8277e-06 - val_loss: 3.9951e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8166e-06 - val_loss: 3.9876e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8060e-06 - val_loss: 3.9947e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.9789e-06 - val_loss: 4.0405e-06
Epoch 2/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.9199e-06 - val_loss: 4.0029e-06
Epoch 3/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8895e-06 - val_loss: 4.0039e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8711e-06 - val_loss: 3.9951e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8537e-06 - val_loss: 4.0026e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8391e-06 - val_loss: 3.9839e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8267e-06 - val_loss: 3.9973e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8155e-06 - val_loss: 3.9859e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8063e-06 - val_loss: 3.9808e-06
Epoch 10/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.7972e-06 - val_loss: 3.9786e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9304e-06 - val_loss: 3.9639e-06
Epoch 2/10
100000/100000 [==============================] - 4s 36us/step - loss: 3.8774e-06 - val_loss: 3.9412e-06
Epoch 3/10
100000/100000 [==============================] - 4s 36us/step - loss: 3.8473e-06 - val_loss: 3.9411e-06
Epoch 4/10
100000/100000 [==============================] - 4s 36us/step - loss: 3.8265e-06 - val_loss: 3.9357e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8117e-06 - val_loss: 3.9413e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7994e-06 - val_loss: 3.9346e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7867e-06 - val_loss: 3.9404e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7767e-06 - val_loss: 3.9429e-06
Epoch 9/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7652e-06 - val_loss: 3.9319e-06
Epoch 10/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7569e-06 - val_loss: 3.9316e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.9209e-06 - val_loss: 3.9507e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.8646e-06 - val_loss: 3.9374e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8370e-06 - val_loss: 3.9381e-06
Epoch 4/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8159e-06 - val_loss: 3.9290e-06
Epoch 5/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.8014e-06 - val_loss: 3.9297e-06
Epoch 6/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7878e-06 - val_loss: 3.9302e-06
Epoch 7/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7772e-06 - val_loss: 3.9285e-06
Epoch 8/10
100000/100000 [==============================] - 3s 35us/step - loss: 3.7653e-06 - val_loss: 3.9249e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7560e-06 - val_loss: 3.9136e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7464e-06 - val_loss: 3.9207e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.8959e-06 - val_loss: 3.9158e-06
Epoch 2/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8418e-06 - val_loss: 3.9084e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8154e-06 - val_loss: 3.9092e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7937e-06 - val_loss: 3.8950e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7804e-06 - val_loss: 3.8992e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7662e-06 - val_loss: 3.8928e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7522e-06 - val_loss: 3.8931e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7433e-06 - val_loss: 3.8963e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7335e-06 - val_loss: 3.8973e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7234e-06 - val_loss: 3.8915e-06
Train on 100000 samples, validate on 20000 samples
Epoch 1/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8911e-06 - val_loss: 3.8913e-06
Epoch 2/10
100000/100000 [==============================] - 4s 35us/step - loss: 3.8377e-06 - val_loss: 3.8628e-06
Epoch 3/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.8092e-06 - val_loss: 3.8597e-06
Epoch 4/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7907e-06 - val_loss: 3.8370e-06
Epoch 5/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7735e-06 - val_loss: 3.8516e-06
Epoch 6/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7603e-06 - val_loss: 3.8562e-06
Epoch 7/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7499e-06 - val_loss: 3.8418e-06
Epoch 8/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7380e-06 - val_loss: 3.8353e-06
Epoch 9/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7278e-06 - val_loss: 3.8471e-06
Epoch 10/10
100000/100000 [==============================] - 3s 34us/step - loss: 3.7188e-06 - val_loss: 3.8480e-06

Once again save the model (although it is already saved each epoch)



In [0]:

    
%cd /gdrive/My Drive/JetImages/QCD
autoencoder.save(ModelName)
# autoencoder = keras.models.load_model(ModelName)



In [0]:

    
%cd /gdrive/My Drive/JetImages/QCD
!ls -lh









    



total 99M
-rw------- 1 root root  52M Nov 29 13:14 Model_40_24_8_24_40_40
-rw------- 1 root root  47M Nov 29 13:14 Model_40_24_8_24_40_40.xz
drwx------ 3 root root 4.0K Nov 28 14:52 TEST
drwx------ 3 root root 4.0K Nov 28 14:52 TRAIN



In [0]:

    
%cd /gdrive/My Drive/JetImages/QCD
!xz -z9evvfk Model_40_24_8_24_40_40









    



/gdrive/My Drive/JetImages/QCD
xz: Filter chain: --lzma2=dict=64MiB,lc=3,lp=0,pb=2,mode=normal,nice=273,mf=bt4,depth=512
xz: 674 MiB of memory is required. The limiter is disabled.
xz: Decompression will need 65 MiB of memory.
Model_40_24_8_24_40_40 (1/1)
  100 %         46.8 MiB / 51.3 MiB = 0.912   1.6 MiB/s       0:32

Evaluate using the trained model:



In [0]:

    
%cd /gdrive/My Drive/JetImages/QCD
EvalOnFile("./TEST/BoxImages/0.xz","./TEST/BoxImages/0_out")
EvalOnFile("./TEST/BoxImages/1.xz","./TEST/BoxImages/1_out")
EvalOnFile("./TEST/BoxImages/2.xz","./TEST/BoxImages/2_out")
EvalOnFile("./TEST/BoxImages/3.xz","./TEST/BoxImages/3_out")
EvalOnFile("./TEST/BoxImages/4.xz","./TEST/BoxImages/4_out")
EvalOnFile("./TEST/BoxImages/5.xz","./TEST/BoxImages/5_out")
EvalOnFile("./TEST/BoxImages/6.xz","./TEST/BoxImages/6_out")
EvalOnFile("./TEST/BoxImages/7.xz","./TEST/BoxImages/7_out")









    



/gdrive/My Drive/JetImages/QCD
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)



In [0]:

    
%cd /gdrive/My Drive/JetImages/TOP
EvalOnFile("./TEST/BoxImages/0.xz","./TEST/BoxImages/0_out")
EvalOnFile("./TEST/BoxImages/1.xz","./TEST/BoxImages/1_out")
EvalOnFile("./TEST/BoxImages/2.xz","./TEST/BoxImages/2_out")
EvalOnFile("./TEST/BoxImages/3.xz","./TEST/BoxImages/3_out")
EvalOnFile("./TEST/BoxImages/4.xz","./TEST/BoxImages/4_out")
EvalOnFile("./TEST/BoxImages/5.xz","./TEST/BoxImages/5_out")
EvalOnFile("./TEST/BoxImages/6.xz","./TEST/BoxImages/6_out")
EvalOnFile("./TEST/BoxImages/7.xz","./TEST/BoxImages/7_out")









    



/gdrive/My Drive/JetImages/TOP
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)
(20000,)



In [0]:

    
%cd /gdrive/My Drive/JetImages/
!cat TOP/TEST/BoxImages/*_out > TOP_OUT
!cat QCD/TEST/BoxImages/*_out > QCD_OUT
!ls TOP/TEST/BoxImages/*_out TOP_OUT -lh
!ls QCD/TEST/BoxImages/*_out QCD_OUT -lh









    



/gdrive/My Drive/JetImages
-rw------- 1 root root 1.3M Nov 30 03:15 TOP_OUT
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/0_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/1_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/2_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/3_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/4_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/5_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/6_out
-rw------- 1 root root 157K Nov 30 03:15 TOP/TEST/BoxImages/7_out
-rw------- 1 root root 1.3M Nov 30 03:15 QCD_OUT
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/0_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/1_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/2_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/3_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/4_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/5_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/6_out
-rw------- 1 root root 157K Nov 30 03:15 QCD/TEST/BoxImages/7_out

Plotting the loss and ROC:



In [0]:

    
%cd /gdrive/My Drive/JetImages/
qcdloss = np.fromfile("QCD_OUT", dtype=float, count=-1, sep='', offset=0)
toploss = np.fromfile("TOP_OUT", dtype=float, count=-1, sep='', offset=0)
qcdloss=np.sort(qcdloss)
toploss=np.sort(toploss)
print(qcdloss.shape)
print(toploss.shape)









    



/gdrive/My Drive/JetImages
(160000,)
(160000,)



In [0]:

    
plt.hist(toploss,100,(0.0,0.4),density=True,histtype='step')
plt.hist(qcdloss,100,(0.0,0.4),density=True,histtype='step')
plt.show()



In [0]:

    
dx = (0.4 - 0.0) / 100.0
qcdeff = np.ones((100))
topeff = np.ones((100))
for i in range(100):
  xval = i*dx
  qcdeff[i]=1.0/(Count(qcdloss,xval)+0.0000000001)
  topeff[i]=Count(toploss,xval)



In [0]:

    
plt.yscale('log')
plt.plot(topeff,qcdeff)









    Out[0]:





[<matplotlib.lines.Line2D at 0x7f1328f71fd0>]



In [0]:

    
%cd /gdrive/My Drive/JetImages/

def ReadLossMass(lossname,massname):
  loss = np.fromfile(lossname, dtype=float, count=-1, sep='', offset=0)
  mass = np.fromfile(massname, dtype='float32', count=-1, sep='', offset=0)
  out = np.ones((mass.shape[0],2))
  for i in range(mass.shape[0]):
    out[i][0] = loss[i]
    out[i][1] = mass[i]
  return out

def GetQCDPair () :
  pair = ReadLossMass("QCD/TEST/BoxImages/0_out","QCD/TEST/Mass/0")
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/1_out","QCD/TEST/Mass/1"),0)
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/2_out","QCD/TEST/Mass/2"),0)
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/3_out","QCD/TEST/Mass/3"),0)
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/4_out","QCD/TEST/Mass/4"),0)
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/5_out","QCD/TEST/Mass/5"),0)
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/6_out","QCD/TEST/Mass/6"),0)
  pair = np.append (pair,ReadLossMass("QCD/TEST/BoxImages/7_out","QCD/TEST/Mass/7"),0)
  return pair

def GetTOPPair () :
  pair = ReadLossMass("TOP/TEST/BoxImages/0_out","TOP/TEST/Mass/0")
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/1_out","TOP/TEST/Mass/1"),0)
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/2_out","TOP/TEST/Mass/2"),0)
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/3_out","TOP/TEST/Mass/3"),0)
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/4_out","TOP/TEST/Mass/4"),0)
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/5_out","TOP/TEST/Mass/5"),0)
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/6_out","TOP/TEST/Mass/6"),0)
  pair = np.append (pair,ReadLossMass("TOP/TEST/BoxImages/7_out","TOP/TEST/Mass/7"),0)
  return pair

qcdpair = GetQCDPair()
toppair = GetTOPPair()









    



/gdrive/My Drive/JetImages

The 2D Histogram of QCD Loss vs Mass



In [0]:

    
#plt.hist(qcdpair[:,1],100,(0.0,300.0),density=True,histtype='step')
#plt.hist(toppair[:,1],100,(0.0,300.0),density=True,histtype='step')
plt.hist2d(qcdpair[:,1],qcdpair[:,0],bins=100,range=[[0,400],[0.0,0.3]])
plt.show()



In [0]:

    
def QCDMassBin(minmass,maxmass):
  ret = np.ones((1))
  for e in range(qcdpair.shape[0]):
    if (minmass < qcdpair[e][1]) and (qcdpair[e][1] < maxmass) :
      if e == 0 :
        ret[e] = qcdpair[e][0]
      else:
        ret = np.append(ret,qcdpair[e][0])
  return ret

plt.hist(QCDMassBin(0,100),100,(0.0,0.4),density=True,histtype='step')
plt.hist(QCDMassBin(100,200),100,(0.0,0.4),density=True,histtype='step')
plt.hist(QCDMassBin(200,300),100,(0.0,0.4),density=True,histtype='step')
plt.hist(QCDMassBin(300,400),100,(0.0,0.4),density=True,histtype='step')
plt.hist(QCDMassBin(400,500),100,(0.0,0.4),density=True,histtype='step')
plt.hist(QCDMassBin(500,5000),100,(0.0,0.4),density=True,histtype='step')
plt.show()