notebook.community

Edit and run



In [1]:

    
import matplotlib.pyplot as plt
%matplotlib inline
import tensorflow as tf
import numpy as np
import pandas as pd
import time

from numpy import genfromtxt
from scipy import stats



In [2]:

    
start_time = time.time()



In [3]:

    
def read_data(file_name):
    df = pd.read_csv(file_name, sep='\t', header=None)
    return df



In [4]:

    
df = read_data('lezhin_public_dataset_training.tsv')









    



C:\Users\deera\Anaconda3\lib\site-packages\IPython\core\interactiveshell.py:2821: DtypeWarning: Columns (7,8) have mixed types. Specify dtype option on import or set low_memory=False.
  if self.run_code(code, result):



In [5]:

    
# df.iloc[:, :20]
del df[7], df[8], df[16], df[18]



In [6]:

    
df.describe()









    Out[6]:







  
    
      
      0
      1
      2
      3
      4
      5
      6
      9
      10
      11
      ...
      167
      168
      169
      170
      171
      172
      173
      174
      175
      176
    
  
  
    
      count
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      ...
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
      1.279027e+06
    
    
      mean
      2.169985e-01
      1.327282e-01
      2.840949e-01
      1.115090e-01
      2.589148e-01
      2.636578e+00
      7.994948e+01
      -5.841105e+14
      1.291841e-02
      1.485504e-05
      ...
      1.166512e-05
      6.843252e-02
      5.184376e-03
      2.199250e-06
      3.106541e-05
      2.997524e-04
      1.249980e-01
      3.696069e-02
      4.047608e-07
      4.232123e-06
    
    
      std
      4.122018e-01
      3.392809e-01
      4.509824e-01
      3.147615e-01
      4.380389e-01
      6.181888e+00
      1.290465e+02
      7.339708e+16
      1.129227e-01
      3.854197e-03
      ...
      1.123908e-03
      1.463250e-01
      3.121712e-02
      4.712688e-04
      2.538336e-03
      8.295776e-03
      2.005742e-01
      7.977981e-02
      1.090358e-04
      8.696757e-04
    
    
      min
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      -9.223372e+18
      0.000000e+00
      0.000000e+00
      ...
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
    
    
      25%
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      1.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      ...
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
    
    
      50%
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      3.000000e+01
      0.000000e+00
      0.000000e+00
      0.000000e+00
      ...
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      2.940000e-02
      0.000000e+00
      0.000000e+00
      0.000000e+00
    
    
      75%
      0.000000e+00
      0.000000e+00
      1.000000e+00
      0.000000e+00
      1.000000e+00
      0.000000e+00
      1.060000e+02
      0.000000e+00
      0.000000e+00
      0.000000e+00
      ...
      0.000000e+00
      4.560000e-02
      0.000000e+00
      0.000000e+00
      0.000000e+00
      0.000000e+00
      1.504000e-01
      2.880000e-02
      0.000000e+00
      0.000000e+00
    
    
      max
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      2.300000e+01
      6.574000e+03
      1.900000e+01
      1.000000e+00
      1.000000e+00
      ...
      3.883000e-01
      8.586000e-01
      9.111000e-01
      4.424000e-01
      5.783000e-01
      8.418000e-01
      9.545000e-01
      7.078000e-01
      1.044000e-01
      4.217000e-01
    
  

8 rows × 173 columns



In [7]:

    
features = df.iloc[:, 1:].values
labels = df.iloc[:, :1].values
print(stats.describe(features).variance)
print(features.shape, labels.shape)









    



[  1.15111543e-01   2.03385131e-01   9.90748072e-02   1.91878069e-01
   3.82157440e+01   1.66529942e+04   5.38713173e+33   1.27515383e-02
   1.48548331e-05   4.08220688e-02   1.02018734e-01   4.10086862e-02
   3.25354759e-03   2.45624487e-01   4.74883876e+01   4.61867511e+01
   6.06847586e+01   6.37896923e+01   3.75022068e+01   3.60551168e+01
   4.33698589e+01   9.70587381e+01   1.86694818e+01   7.72218115e+01
   2.60344947e+01   1.45445092e+01   1.19066626e+01   3.46125399e+01
   6.31201610e+01   3.17641667e+01   7.50052257e+01   1.10501155e+01
   2.10309402e+01   1.66279723e+01   7.25443018e+01   1.77452340e+01
   1.15795158e+01   4.99343761e+01   2.28384809e+01   3.34100286e+01
   6.02445649e+01   2.68082260e+01   4.05038849e+01   4.55097056e+01
   6.09638438e+01   8.83403558e+00   2.79942773e+01   6.37632101e+01
   3.38823571e+01   4.19303788e+01   3.53614873e+01   5.84163897e+00
   9.37152581e+00   2.26425774e+01   7.53624179e+00   1.58830375e+00
   7.40874098e+00   8.91352938e+00   5.67670481e+01   3.15646487e+01
   1.88409530e+01   2.97533790e+00   2.57013369e+01   8.69676607e+01
   1.13740192e+01   3.09682231e+00   1.11995316e+01   1.44935524e+01
   3.65948879e+01   9.05946851e+00   1.80243176e+01   4.53505973e+01
   4.19139430e+01   5.24059188e+01   1.77892739e+01   6.11265476e+00
   6.15705309e+01   5.78428819e+00   4.70996392e+01   2.05796618e+00
   1.72203324e+01   2.59883616e+01   1.31880166e+01   6.29891644e+01
   1.17708068e+01   6.87302523e+00   3.32261266e+00   3.14228246e+01
   2.40979133e+01   3.67074374e+01   6.34521265e+00   1.94595683e+00
   8.96840824e+00   3.70723767e+00   2.40581581e+00   1.31997267e+01
   5.72881828e+00   7.37273943e+00   6.55531264e+00   2.52547546e+00
   2.83779588e+01   4.37881850e+00   5.84959458e+00   2.36336898e+01
   1.80687010e+01   2.37197825e+00   2.56651371e+01   7.77274626e+00
   2.72595326e+00   2.12897902e+01   3.37070324e+01   1.72273573e+01
   2.48940754e+01   6.63103488e+00   1.61035545e-01   4.53701783e+00
   1.08199063e-01   2.58940862e+01   6.28874443e-02   6.22443148e-02
   8.11025826e-02   6.53844350e-02   6.45557641e-02   5.79377248e-02
   7.79227940e-02   7.54706442e-02   9.43576661e-04   2.40750251e-04
   4.55544444e-02   1.42557805e-01   1.31552615e-01   5.86825921e-02
   1.17556638e-01   8.89872657e-02   2.89534020e-03   9.71820788e-02
   2.44651245e-02   8.53862347e-03   1.67008479e-02   3.20771776e-03
   1.56981807e-03   2.60514984e-02   1.54052207e-01   9.78427646e-02
   1.92542176e-02   2.28867091e-03   7.70755513e+01   6.82408384e-02
   5.75520929e+01   1.40888257e+01   0.00000000e+00   0.00000000e+00
   9.44522435e-03   2.50892224e-02   6.79125738e-03   8.12044538e-03
   6.93538691e-03   8.24447572e-02   4.55275512e-02   4.48746349e-04
   3.53589387e-03   7.55808326e-03   1.26316914e-06   2.14110115e-02
   9.74508785e-04   2.22094237e-07   6.44314914e-06   6.88199069e-05
   4.02300085e-02   6.36481855e-03   1.18887970e-08   7.56335768e-07]
(1279027, 172) (1279027, 1)



In [8]:

    
rnd_indices = np.random.rand(len(features)) < 0.70

train_x = features[rnd_indices]
train_y = labels[rnd_indices]
test_x = features[~rnd_indices]
test_y = labels[~rnd_indices]
print("train row count : %d, test row count : %d" % (train_x.shape[0], test_x.shape[0]))

feature_count = train_x.shape[1]
label_count = train_y.shape[1]
print(feature_count, label_count)









    



train row count : 896439, test row count : 382588
172 1



In [9]:

    
training_epochs = 90
learning_rate = 0.01
cost_history = np.empty(shape=[1],dtype=float)
nb_classes = 2

X = tf.placeholder(tf.float32,[None,feature_count])
Y = tf.placeholder(tf.int32,[None,label_count])
Y_one_hot = tf.one_hot(Y, nb_classes)  # one hot
print("one_hot", Y_one_hot)
Y_one_hot = tf.reshape(Y_one_hot, [-1, nb_classes])
print("reshape", Y_one_hot)









    



one_hot Tensor("one_hot:0", shape=(?, 1, 2), dtype=float32)
reshape Tensor("Reshape:0", shape=(?, 2), dtype=float32)



In [10]:

    
def init_weights(shape):
    return tf.Variable(tf.random_normal(shape, stddev=0.1)), tf.Variable(tf.random_normal([shape[1]]))

def make_hidden_layer(previous_h, weight, bias, p_keep_hidden, is_dropout=True):
    h = tf.nn.relu(tf.matmul(previous_h, weight) + bias)
    if is_dropout:
        h = tf.nn.dropout(h, p_keep_hidden)
    return h

def model(X, p_keep_hidden):
    s_1 = feature_count + 2
    s_2 = feature_count + 2
    s_3 = feature_count
    
    w_h, b = init_weights([feature_count, s_1])
    w_h2, b2 = init_weights([s_1, s_2])
    w_h3, b3 = init_weights([s_2, s_3])
    w_o, b_o = init_weights([s_3, nb_classes])
    
    h = make_hidden_layer(X, w_h, b, p_keep_hidden)
    h2 = make_hidden_layer(h, w_h2, b2, p_keep_hidden)
    h3 = make_hidden_layer(h2, w_h3, b3, p_keep_hidden, False)
    
    return tf.matmul(h3, w_o) + b_o



In [11]:

    
p_keep_hidden = tf.placeholder("float")

h0 = model(X, p_keep_hidden)



In [12]:

    
# Cross entropy cost/loss
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=h0, labels=Y_one_hot))
optimizer = tf.train.AdamOptimizer(learning_rate).minimize(cost)



In [13]:

    
prediction = tf.argmax(h0, 1)
correct_prediction = tf.equal(prediction, tf.argmax(Y_one_hot, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))



In [14]:

    
print(train_x.shape, train_y.shape)
print(test_x.shape, test_y.shape)
print(X.shape, Y.shape)
training_dropout_h = 0.95

batch_size = 2000
batch_length = int(train_x.shape[0] / batch_size)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())

    for step in range(training_epochs + 1):
        for batch_num in range(batch_length):
            start_idx = batch_num * batch_size
            end_idx = (train_x.shape[0] - 1) if batch_num == batch_length - 1 else (batch_num + 1) * batch_size
                
            if batch_num % 200 == 0 or batch_num == batch_length - 1:
                print("batch num : %d / %d, index: %d ~ %d" % (batch_num, batch_length - 1, start_idx, end_idx))
            
            sess.run(optimizer, feed_dict={X: train_x[start_idx:end_idx], Y: train_y[start_idx:end_idx], p_keep_hidden: training_dropout_h})

        loss, acc = sess.run([cost, accuracy], feed_dict={
                                 X: train_x, Y: train_y, p_keep_hidden: training_dropout_h})
        cost_history = np.append(cost_history, acc)
        if step % 4 == 0:
            print("Step: {:5}\tLoss: {:.3f}\tAcc: {:.2%}".format(
                step, loss, acc))
            
    # Test model and check accuracy
    pre = tf.argmax(h0, 1)
    test_yy = np.transpose(test_y.ravel())
    print(test_yy.shape)
    correct_prediction = tf.equal(pre, test_yy)
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    print('Test Accuracy:', sess.run(accuracy, feed_dict={X: test_x, p_keep_hidden: 1.0}))









    



(896439, 172) (896439, 1)
(382588, 172) (382588, 1)
(?, 172) (?, 1)
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:     0	Loss: 1.465	Acc: 51.58%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:     4	Loss: 0.354	Acc: 88.22%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:     8	Loss: 0.360	Acc: 88.03%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    12	Loss: 0.114	Acc: 97.67%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    16	Loss: 0.360	Acc: 88.91%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    20	Loss: 0.215	Acc: 93.16%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    24	Loss: 0.207	Acc: 94.57%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    28	Loss: 0.219	Acc: 93.82%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    32	Loss: 0.065	Acc: 98.10%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    36	Loss: 0.033	Acc: 98.91%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    40	Loss: 0.026	Acc: 99.07%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    44	Loss: 0.024	Acc: 99.29%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    48	Loss: 0.021	Acc: 99.27%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    52	Loss: 0.022	Acc: 99.24%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    56	Loss: 0.019	Acc: 99.47%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    60	Loss: 0.014	Acc: 99.59%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    64	Loss: 0.007	Acc: 99.85%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    68	Loss: 0.006	Acc: 99.89%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    72	Loss: 0.008	Acc: 99.86%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    76	Loss: 0.009	Acc: 99.82%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    80	Loss: 0.010	Acc: 99.84%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    84	Loss: 0.004	Acc: 99.91%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
Step:    88	Loss: 0.005	Acc: 99.88%
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
batch num : 0 / 447, index: 0 ~ 2000
batch num : 200 / 447, index: 400000 ~ 402000
batch num : 400 / 447, index: 800000 ~ 802000
batch num : 447 / 447, index: 894000 ~ 896438
(382588,)
Test Accuracy: 0.999856



In [15]:

    
print(cost_history.shape)
plt.plot(range(len(cost_history)),cost_history)
plt.axis([0,training_epochs,0,1])
plt.show()









    



(92,)



In [16]:

    
sess.close()
end_time = time.time()
print("processing time : %d seconds" % (end_time - start_time,))









    



processing time : 241 seconds

	0	1	2	3	4	5	6	9	10	11	...	167	168	169	170	171	172	173	174	175	176
count	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	...	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06	1.279027e+06
mean	2.169985e-01	1.327282e-01	2.840949e-01	1.115090e-01	2.589148e-01	2.636578e+00	7.994948e+01	-5.841105e+14	1.291841e-02	1.485504e-05	...	1.166512e-05	6.843252e-02	5.184376e-03	2.199250e-06	3.106541e-05	2.997524e-04	1.249980e-01	3.696069e-02	4.047608e-07	4.232123e-06
std	4.122018e-01	3.392809e-01	4.509824e-01	3.147615e-01	4.380389e-01	6.181888e+00	1.290465e+02	7.339708e+16	1.129227e-01	3.854197e-03	...	1.123908e-03	1.463250e-01	3.121712e-02	4.712688e-04	2.538336e-03	8.295776e-03	2.005742e-01	7.977981e-02	1.090358e-04	8.696757e-04
min	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	-9.223372e+18	0.000000e+00	0.000000e+00	...	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00
25%	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	1.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	...	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00
50%	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	3.000000e+01	0.000000e+00	0.000000e+00	0.000000e+00	...	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	2.940000e-02	0.000000e+00	0.000000e+00	0.000000e+00
75%	0.000000e+00	0.000000e+00	1.000000e+00	0.000000e+00	1.000000e+00	0.000000e+00	1.060000e+02	0.000000e+00	0.000000e+00	0.000000e+00	...	0.000000e+00	4.560000e-02	0.000000e+00	0.000000e+00	0.000000e+00	0.000000e+00	1.504000e-01	2.880000e-02	0.000000e+00	0.000000e+00
max	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	2.300000e+01	6.574000e+03	1.900000e+01	1.000000e+00	1.000000e+00	...	3.883000e-01	8.586000e-01	9.111000e-01	4.424000e-01	5.783000e-01	8.418000e-01	9.545000e-01	7.078000e-01	1.044000e-01	4.217000e-01