In [ ]:

    

'''
A Convolutional Network implementation example using TensorFlow library.
This example is using the MNIST database of handwritten digits
(http://yann.lecun.com/exdb/mnist/)

Author: Aymeric Damien
Project: https://github.com/aymericdamien/TensorFlow-Examples/
'''


    

In [ ]:

    

import tensorflow as tf

# Import MNIST data
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)


    

In [2]:

    

# Parameters
learning_rate = 0.001
training_iters = 200000
batch_size = 128
display_step = 10

# Network Parameters
n_input = 784 # MNIST data input (img shape: 28*28)
n_classes = 10 # MNIST total classes (0-9 digits)
dropout = 0.75 # Dropout, probability to keep units

# tf Graph input
x = tf.placeholder(tf.float32, [None, n_input])
y = tf.placeholder(tf.float32, [None, n_classes])
keep_prob = tf.placeholder(tf.float32) #dropout (keep probability)


    

In [3]:

    

# Create some wrappers for simplicity
def conv2d(x, W, b, strides=1):
    # Conv2D wrapper, with bias and relu activation
    x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')
    x = tf.nn.bias_add(x, b)
    return tf.nn.relu(x)


def maxpool2d(x, k=2):
    # MaxPool2D wrapper
    return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1],
                          padding='SAME')


# Create model
def conv_net(x, weights, biases, dropout):
    # Reshape input picture
    x = tf.reshape(x, shape=[-1, 28, 28, 1])

    # Convolution Layer
    conv1 = conv2d(x, weights['wc1'], biases['bc1'])
    # Max Pooling (down-sampling)
    conv1 = maxpool2d(conv1, k=2)

    # Convolution Layer
    conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])
    # Max Pooling (down-sampling)
    conv2 = maxpool2d(conv2, k=2)

    # Fully connected layer
    # Reshape conv2 output to fit fully connected layer input
    fc1 = tf.reshape(conv2, [-1, weights['wd1'].get_shape().as_list()[0]])
    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])
    fc1 = tf.nn.relu(fc1)
    # Apply Dropout
    fc1 = tf.nn.dropout(fc1, dropout)

    # Output, class prediction
    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])
    return out


    

In [4]:

    

# Store layers weight & bias
weights = {
    # 5x5 conv, 1 input, 32 outputs
    'wc1': tf.Variable(tf.random_normal([5, 5, 1, 32])),
    # 5x5 conv, 32 inputs, 64 outputs
    'wc2': tf.Variable(tf.random_normal([5, 5, 32, 64])),
    # fully connected, 7*7*64 inputs, 1024 outputs
    'wd1': tf.Variable(tf.random_normal([7*7*64, 1024])),
    # 1024 inputs, 10 outputs (class prediction)
    'out': tf.Variable(tf.random_normal([1024, n_classes]))
}

biases = {
    'bc1': tf.Variable(tf.random_normal([32])),
    'bc2': tf.Variable(tf.random_normal([64])),
    'bd1': tf.Variable(tf.random_normal([1024])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}

# Construct model
pred = conv_net(x, weights, biases, keep_prob)

# Define loss and optimizer
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)

# Evaluate model
correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# Initializing the variables
init = tf.global_variables_initializer()


    

In [5]:

    

# Launch the graph
with tf.Session() as sess:
    sess.run(init)
    step = 1
    # Keep training until reach max iterations
    while step * batch_size < training_iters:
        batch_x, batch_y = mnist.train.next_batch(batch_size)
        # Run optimization op (backprop)
        sess.run(optimizer, feed_dict={x: batch_x, y: batch_y,
                                       keep_prob: dropout})
        if step % display_step == 0:
            # Calculate batch loss and accuracy
            loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_x,
                                                              y: batch_y,
                                                              keep_prob: 1.})
            print("Iter " + str(step*batch_size) + ", Minibatch Loss= " + \
                  "{:.6f}".format(loss) + ", Training Accuracy= " + \
                  "{:.5f}".format(acc))
        step += 1
    print("Optimization Finished!")

    # Calculate accuracy for 256 mnist test images
    print("Testing Accuracy:", \
        sess.run(accuracy, feed_dict={x: mnist.test.images[:256],
                                      y: mnist.test.labels[:256],
                                      keep_prob: 1.}))


    

    




Iter 1280, Minibatch Loss= 26574.855469, Training Accuracy= 0.25781
Iter 2560, Minibatch Loss= 11454.494141, Training Accuracy= 0.49219
Iter 3840, Minibatch Loss= 10070.515625, Training Accuracy= 0.55469
Iter 5120, Minibatch Loss= 4008.586426, Training Accuracy= 0.78125
Iter 6400, Minibatch Loss= 3148.004639, Training Accuracy= 0.80469
Iter 7680, Minibatch Loss= 6740.440430, Training Accuracy= 0.71875
Iter 8960, Minibatch Loss= 4103.991699, Training Accuracy= 0.80469
Iter 10240, Minibatch Loss= 2631.275391, Training Accuracy= 0.85938
Iter 11520, Minibatch Loss= 1428.798828, Training Accuracy= 0.91406
Iter 12800, Minibatch Loss= 3909.772705, Training Accuracy= 0.78906
Iter 14080, Minibatch Loss= 1423.095947, Training Accuracy= 0.88281
Iter 15360, Minibatch Loss= 1524.569824, Training Accuracy= 0.89062
Iter 16640, Minibatch Loss= 2234.539795, Training Accuracy= 0.86719
Iter 17920, Minibatch Loss= 933.932800, Training Accuracy= 0.90625
Iter 19200, Minibatch Loss= 2039.046021, Training Accuracy= 0.89062
Iter 20480, Minibatch Loss= 674.179932, Training Accuracy= 0.95312
Iter 21760, Minibatch Loss= 3778.958984, Training Accuracy= 0.82812
Iter 23040, Minibatch Loss= 1038.217773, Training Accuracy= 0.91406
Iter 24320, Minibatch Loss= 1689.513672, Training Accuracy= 0.89062
Iter 25600, Minibatch Loss= 1800.954956, Training Accuracy= 0.85938
Iter 26880, Minibatch Loss= 1086.292847, Training Accuracy= 0.90625
Iter 28160, Minibatch Loss= 656.042847, Training Accuracy= 0.94531
Iter 29440, Minibatch Loss= 1210.589844, Training Accuracy= 0.91406
Iter 30720, Minibatch Loss= 1099.606323, Training Accuracy= 0.90625
Iter 32000, Minibatch Loss= 1073.128174, Training Accuracy= 0.92969
Iter 33280, Minibatch Loss= 518.844543, Training Accuracy= 0.95312
Iter 34560, Minibatch Loss= 540.856689, Training Accuracy= 0.92188
Iter 35840, Minibatch Loss= 353.990906, Training Accuracy= 0.97656
Iter 37120, Minibatch Loss= 1488.962891, Training Accuracy= 0.91406
Iter 38400, Minibatch Loss= 231.191864, Training Accuracy= 0.98438
Iter 39680, Minibatch Loss= 171.154480, Training Accuracy= 0.98438
Iter 40960, Minibatch Loss= 2092.023682, Training Accuracy= 0.90625
Iter 42240, Minibatch Loss= 480.594299, Training Accuracy= 0.95312
Iter 43520, Minibatch Loss= 504.128143, Training Accuracy= 0.96875
Iter 44800, Minibatch Loss= 143.534485, Training Accuracy= 0.97656
Iter 46080, Minibatch Loss= 325.875580, Training Accuracy= 0.96094
Iter 47360, Minibatch Loss= 602.813049, Training Accuracy= 0.91406
Iter 48640, Minibatch Loss= 794.595093, Training Accuracy= 0.94531
Iter 49920, Minibatch Loss= 415.539032, Training Accuracy= 0.95312
Iter 51200, Minibatch Loss= 146.016022, Training Accuracy= 0.96094
Iter 52480, Minibatch Loss= 294.180786, Training Accuracy= 0.94531
Iter 53760, Minibatch Loss= 50.955730, Training Accuracy= 0.99219
Iter 55040, Minibatch Loss= 1026.607056, Training Accuracy= 0.92188
Iter 56320, Minibatch Loss= 283.756134, Training Accuracy= 0.96875
Iter 57600, Minibatch Loss= 691.538208, Training Accuracy= 0.95312
Iter 58880, Minibatch Loss= 491.075073, Training Accuracy= 0.96094
Iter 60160, Minibatch Loss= 571.951660, Training Accuracy= 0.95312
Iter 61440, Minibatch Loss= 284.041168, Training Accuracy= 0.97656
Iter 62720, Minibatch Loss= 1041.941528, Training Accuracy= 0.92969
Iter 64000, Minibatch Loss= 664.833923, Training Accuracy= 0.93750
Iter 65280, Minibatch Loss= 1582.112793, Training Accuracy= 0.88281
Iter 66560, Minibatch Loss= 783.135376, Training Accuracy= 0.94531
Iter 67840, Minibatch Loss= 245.942398, Training Accuracy= 0.96094
Iter 69120, Minibatch Loss= 752.858948, Training Accuracy= 0.96875
Iter 70400, Minibatch Loss= 623.243286, Training Accuracy= 0.94531
Iter 71680, Minibatch Loss= 846.498230, Training Accuracy= 0.93750
Iter 72960, Minibatch Loss= 586.516479, Training Accuracy= 0.95312
Iter 74240, Minibatch Loss= 92.774963, Training Accuracy= 0.98438
Iter 75520, Minibatch Loss= 644.039612, Training Accuracy= 0.95312
Iter 76800, Minibatch Loss= 693.247681, Training Accuracy= 0.96094
Iter 78080, Minibatch Loss= 466.491882, Training Accuracy= 0.96094
Iter 79360, Minibatch Loss= 964.212341, Training Accuracy= 0.93750
Iter 80640, Minibatch Loss= 230.451904, Training Accuracy= 0.97656
Iter 81920, Minibatch Loss= 280.434570, Training Accuracy= 0.95312
Iter 83200, Minibatch Loss= 213.208252, Training Accuracy= 0.97656
Iter 84480, Minibatch Loss= 774.836060, Training Accuracy= 0.94531
Iter 85760, Minibatch Loss= 164.687729, Training Accuracy= 0.96094
Iter 87040, Minibatch Loss= 419.967407, Training Accuracy= 0.96875
Iter 88320, Minibatch Loss= 160.920151, Training Accuracy= 0.96875
Iter 89600, Minibatch Loss= 586.063599, Training Accuracy= 0.96094
Iter 90880, Minibatch Loss= 345.598145, Training Accuracy= 0.96875
Iter 92160, Minibatch Loss= 931.361145, Training Accuracy= 0.92188
Iter 93440, Minibatch Loss= 170.107117, Training Accuracy= 0.97656
Iter 94720, Minibatch Loss= 497.162750, Training Accuracy= 0.93750
Iter 96000, Minibatch Loss= 906.600464, Training Accuracy= 0.94531
Iter 97280, Minibatch Loss= 303.382202, Training Accuracy= 0.92969
Iter 98560, Minibatch Loss= 509.161652, Training Accuracy= 0.97656
Iter 99840, Minibatch Loss= 359.561981, Training Accuracy= 0.97656
Iter 101120, Minibatch Loss= 136.516541, Training Accuracy= 0.97656
Iter 102400, Minibatch Loss= 517.199341, Training Accuracy= 0.96875
Iter 103680, Minibatch Loss= 487.793335, Training Accuracy= 0.95312
Iter 104960, Minibatch Loss= 407.351929, Training Accuracy= 0.96094
Iter 106240, Minibatch Loss= 70.495193, Training Accuracy= 0.98438
Iter 107520, Minibatch Loss= 344.783508, Training Accuracy= 0.96094
Iter 108800, Minibatch Loss= 242.682465, Training Accuracy= 0.95312
Iter 110080, Minibatch Loss= 169.181458, Training Accuracy= 0.96094
Iter 111360, Minibatch Loss= 152.638245, Training Accuracy= 0.98438
Iter 112640, Minibatch Loss= 170.795868, Training Accuracy= 0.96875
Iter 113920, Minibatch Loss= 133.262726, Training Accuracy= 0.98438
Iter 115200, Minibatch Loss= 296.063293, Training Accuracy= 0.95312
Iter 116480, Minibatch Loss= 254.247543, Training Accuracy= 0.96094
Iter 117760, Minibatch Loss= 506.795715, Training Accuracy= 0.94531
Iter 119040, Minibatch Loss= 446.006897, Training Accuracy= 0.96094
Iter 120320, Minibatch Loss= 149.467377, Training Accuracy= 0.97656
Iter 121600, Minibatch Loss= 52.783600, Training Accuracy= 0.98438
Iter 122880, Minibatch Loss= 49.041794, Training Accuracy= 0.98438
Iter 124160, Minibatch Loss= 184.371246, Training Accuracy= 0.97656
Iter 125440, Minibatch Loss= 129.838501, Training Accuracy= 0.97656
Iter 126720, Minibatch Loss= 288.006531, Training Accuracy= 0.96875
Iter 128000, Minibatch Loss= 187.284653, Training Accuracy= 0.97656
Iter 129280, Minibatch Loss= 197.969955, Training Accuracy= 0.96875
Iter 130560, Minibatch Loss= 299.969818, Training Accuracy= 0.96875
Iter 131840, Minibatch Loss= 537.602173, Training Accuracy= 0.96094
Iter 133120, Minibatch Loss= 4.519302, Training Accuracy= 0.99219
Iter 134400, Minibatch Loss= 133.264191, Training Accuracy= 0.97656
Iter 135680, Minibatch Loss= 89.662292, Training Accuracy= 0.97656
Iter 136960, Minibatch Loss= 107.774078, Training Accuracy= 0.96875
Iter 138240, Minibatch Loss= 335.904572, Training Accuracy= 0.96094
Iter 139520, Minibatch Loss= 457.494568, Training Accuracy= 0.96094
Iter 140800, Minibatch Loss= 259.131531, Training Accuracy= 0.95312
Iter 142080, Minibatch Loss= 152.205383, Training Accuracy= 0.96094
Iter 143360, Minibatch Loss= 252.535828, Training Accuracy= 0.95312
Iter 144640, Minibatch Loss= 109.477585, Training Accuracy= 0.96875
Iter 145920, Minibatch Loss= 24.468613, Training Accuracy= 0.99219
Iter 147200, Minibatch Loss= 51.722107, Training Accuracy= 0.97656
Iter 148480, Minibatch Loss= 69.715233, Training Accuracy= 0.97656
Iter 149760, Minibatch Loss= 405.289246, Training Accuracy= 0.92969
Iter 151040, Minibatch Loss= 282.976379, Training Accuracy= 0.95312
Iter 152320, Minibatch Loss= 134.991119, Training Accuracy= 0.97656
Iter 153600, Minibatch Loss= 491.618103, Training Accuracy= 0.92188
Iter 154880, Minibatch Loss= 154.299988, Training Accuracy= 0.99219
Iter 156160, Minibatch Loss= 79.480019, Training Accuracy= 0.96875
Iter 157440, Minibatch Loss= 68.093750, Training Accuracy= 0.99219
Iter 158720, Minibatch Loss= 459.739685, Training Accuracy= 0.92188
Iter 160000, Minibatch Loss= 168.076843, Training Accuracy= 0.94531
Iter 161280, Minibatch Loss= 256.141846, Training Accuracy= 0.97656
Iter 162560, Minibatch Loss= 236.400391, Training Accuracy= 0.94531
Iter 163840, Minibatch Loss= 177.011261, Training Accuracy= 0.96875
Iter 165120, Minibatch Loss= 48.583298, Training Accuracy= 0.97656
Iter 166400, Minibatch Loss= 413.800293, Training Accuracy= 0.96094
Iter 167680, Minibatch Loss= 209.587387, Training Accuracy= 0.96875
Iter 168960, Minibatch Loss= 239.407318, Training Accuracy= 0.98438
Iter 170240, Minibatch Loss= 183.567017, Training Accuracy= 0.96875
Iter 171520, Minibatch Loss= 87.937515, Training Accuracy= 0.96875
Iter 172800, Minibatch Loss= 203.777039, Training Accuracy= 0.98438
Iter 174080, Minibatch Loss= 566.378052, Training Accuracy= 0.94531
Iter 175360, Minibatch Loss= 325.170898, Training Accuracy= 0.95312
Iter 176640, Minibatch Loss= 300.142212, Training Accuracy= 0.97656
Iter 177920, Minibatch Loss= 205.370193, Training Accuracy= 0.95312
Iter 179200, Minibatch Loss= 5.594437, Training Accuracy= 0.99219
Iter 180480, Minibatch Loss= 110.732109, Training Accuracy= 0.98438
Iter 181760, Minibatch Loss= 33.320297, Training Accuracy= 0.99219
Iter 183040, Minibatch Loss= 6.885544, Training Accuracy= 0.99219
Iter 184320, Minibatch Loss= 221.144806, Training Accuracy= 0.96875
Iter 185600, Minibatch Loss= 365.337372, Training Accuracy= 0.94531
Iter 186880, Minibatch Loss= 186.558258, Training Accuracy= 0.96094
Iter 188160, Minibatch Loss= 149.720322, Training Accuracy= 0.98438
Iter 189440, Minibatch Loss= 105.281998, Training Accuracy= 0.97656
Iter 190720, Minibatch Loss= 289.980011, Training Accuracy= 0.96094
Iter 192000, Minibatch Loss= 214.382278, Training Accuracy= 0.96094
Iter 193280, Minibatch Loss= 461.044312, Training Accuracy= 0.93750
Iter 194560, Minibatch Loss= 138.653076, Training Accuracy= 0.98438
Iter 195840, Minibatch Loss= 112.004883, Training Accuracy= 0.98438
Iter 197120, Minibatch Loss= 212.691467, Training Accuracy= 0.97656
Iter 198400, Minibatch Loss= 57.642502, Training Accuracy= 0.97656
Iter 199680, Minibatch Loss= 80.503563, Training Accuracy= 0.96875
Optimization Finished!
Testing Accuracy: 0.984375

In [ ]: