https://www.tensorflow.org/get_started/mnist/pros
In [1]:
#load mnist data
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
We're going to eventually define a graph which will represent a "dataflow computation". Before we start buiding our graph by creating nodes, we first initial a tf.session. A session allows us to execute graphs. It also allows for the specification of resource allocation (more than one CPU/GPU/machine). The session also holds the values of our intermediate results during training and the values of variables during training.
In [3]:
#Start TensorFlow InteractiveSession
import tensorflow as tf
sess = tf.InteractiveSession()
In [4]:
# create placeholder nodes for the input images and target output
#x will consist of a 2d tensor floating point numbers. 784 = 28*28 pixels
# None indicates the batch size, because we specify 'None' can be a variable size
x = tf.placeholder(tf.float32, shape=[None, 784])
#y_ is another 2d tensor, where each row isicallly a one-hot 10 diminsional vector
#shape option allows TF to automatically catch bugs due to inconsistent
# tensor shapes
y_ = tf.placeholder(tf.float32, shape=[None, 10])
In [7]:
W = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))
Before we can use them, gotta intialize them
In [8]:
sess.run(tf.global_variables_initializer())
Let's add in a regression model.
In [9]:
#x: input images
#W: weight matrix
#b: bias
y = tf.matmul(x,W) + b
Specify a corss-entropy loss function. So, the cross-entropy between the target and the softmax activation function applied to the model's prediction.
Note that tf.nn.softmax_cross_entropy_with_logits internally applies the softmax on the model's unnormalized model prediction and sums across all classes, and tf.reduce_mean takes the average over these sums.
In [11]:
cross_entropy = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y))
In [15]:
%%time
train_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)
for _ in range(1000):
batch = mnist.train.next_batch(100)
train_step.run(feed_dict={x: batch[0], y_: batch[1]})
In [15]:
batch = mnist.train.next_batch(100)
batch[1].shape
Out[15]:
In [20]:
#tf.argmax gives an index of the highest entry in a tensor along some axis
correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))
#we can take this list of booleans and calculate the fraction correct
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
#print the accuracy
print(accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels}))
In [21]:
#because we're gonna need
def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial)
def bias_variable(shape):
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial)
In [24]:
def conv2d(x, W):
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
strides=[1, 2, 2, 1], padding='SAME')
In [25]:
#first convo layer
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
#reshape x to a 4d tensor
x_image = tf.reshape(x, [-1,28,28,1])
#convolve x_image with the weight tensor, add bias,
#apply the ReLU function, and finally max pool
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
In [26]:
#second convo layer
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
In [27]:
# densely connected layer
W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])
h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)
In [28]:
# dropout
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
In [29]:
# readout layer
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
In [30]:
%% time
cross_entropy = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
sess.run(tf.global_variables_initializer())
for i in range(20000):
batch = mnist.train.next_batch(50)
if i%100 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:batch[0], y_: batch[1], keep_prob: 1.0})
print("step %d, training accuracy %g"%(i, train_accuracy))
train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})
print("test accuracy %g"%accuracy.eval(feed_dict={
x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))
In [ ]:
In [ ]: