MNIST: x is just a placeholder ，来自MNIST的数据如何进入占位符x？

Question

这是 senddex 教程中的代码：来自 MNIST 集的数据如何传输到占位符 x.

请帮助我，考虑到我只是 tensorflow 的初学者，如果它与占位符有关，请解释。

提前致谢！

"""
os.environ removes the warning
"""
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
"""
tensorflow starts below
"""

import tensorflow as tf

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/",one_hot=True)

# 10 classes , 0-9

"""
nodes for the hidden layers
"""
n_nodes_hl1 = 500
n_nodes_hl2 = 500
n_nodes_hl3 = 500

n_classes = 10 # 0-9

batch_size = 100

"""
placeholders
"""

x = tf.placeholder('float',[None,784]) # 784 is 28*28 ,i.e., the size of mnist images
y = tf.placeholder('float')


# y is the label of data

def neural_network_model(data):


    # biases are added so that the some neurons get fired even when input_data is 0

    hidden_1_layer = {'weights':tf.Variable(tf.random_normal([784,n_nodes_hl1])),'biases':tf.Variable(tf.random_normal([n_nodes_hl1]))}
    hidden_2_layer = {'weights':tf.Variable(tf.random_normal([n_nodes_hl1,n_nodes_hl2])),'biases':tf.Variable(tf.random_normal([n_nodes_hl2]))}
    hidden_3_layer = {'weights':tf.Variable(tf.random_normal([n_nodes_hl2,n_nodes_hl3])),'biases':tf.Variable(tf.random_normal([n_nodes_hl3]))}

    output_layer = {'weights':tf.Variable(tf.random_normal([n_nodes_hl3,n_classes])),
    'biases':tf.Variable(tf.random_normal([n_classes]))}


    # (input_data * weights) + biases

    l1 = tf.add(tf.matmul(data,hidden_1_layer['weights']) , hidden_1_layer['biases'])
    l1 = tf.nn.relu(l1) # activation func

    l2 = tf.add(tf.matmul(l1,hidden_2_layer['weights']) , hidden_2_layer['biases'])
    l2 = tf.nn.relu(l2) # activation func

    l3 = tf.add(tf.matmul(l2,hidden_3_layer['weights']) , hidden_3_layer['biases'])
    l3 = tf.nn.relu(l3) # activation func

    output = tf.matmul(l3,output_layer['weights']) + output_layer['biases'] 

    return output

# we now have modeled a neural network

def train_neural_network(x):
    prediction = neural_network_model(x)
    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,labels=y))
    # softmax_cross_entropy_with_logits ==> for changing weights
    # we wanna minimize the difference

    # AdamOptimizer optionally has a learning_reate : 0.0001
    optimizer = tf.train.AdamOptimizer().minimize(cost)

    hm_epochs = 5 # cycles of feed forward + back

    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer()) # replace it with global_variable_initializer

        for epoch in range(hm_epochs):
            epoch_loss = 0
            for _ in range(int(mnist.train.num_examples/batch_size)):
                epoch_x,epoch_y = mnist.train.next_batch(batch_size) 
                _,c = sess.run([optimizer, cost], feed_dict = {x: epoch_x, y: epoch_y})
                epoch_loss += c
            print('Epoch',epoch,'completed out of',hm_epochs,' loss:',epoch_loss)


        correct = tf.equal(tf.argmax(prediction,1),tf.argmax(y,1))
        accuracy = tf.reduce_mean(tf.cast(correct,'float')) # cast changes the data type of a tensor
        print('Accuracy: ',accuracy.eval({x:mnist.test.images,y:mnist.test.labels}))


if __name__ == "__main__":
    train_neural_network(x)

Answer 1

要查看 MNIST 数据在何处传输到 tf.placeholder() 张量 x 和 y，请关注这些行：

for _ in range(int(mnist.train.num_examples/batch_size)):
    epoch_x, epoch_y = mnist.train.next_batch(batch_size) 
    _, c = sess.run([optimizer, cost], feed_dict = {x: epoch_x, y: epoch_y})

数组 epoch_x 和 epoch_y 是一对（命名有点混乱）NumPy 数组，分别包含来自 MNIST batch_size 的一批图像和标签 training 数据集。它们将在 for 循环的每次迭代中包含不同的批次。

sess.run() 的 feed_dict 参数告诉 TensorFlow 将 epoch_x 的值替换为占位符 x，将 epoch_y 的值替换为占位符 y。因此，TensorFlow 将使用这些值运行优化算法的一个步骤（在本例中为 Adam）。

注意这一行也用到了MNIST数据：

print('Accuracy: ', accuracy.eval({x: mnist.test.images, y: mnist.test.labels}))

...除了此处，程序使用整个 test 数据集来评估模型的准确性。

MNIST: x is just a placeholder ，来自MNIST的数据如何进入占位符x？

MNIST: x is just a placeholder , How is data from MNIST going into the placeholder x?

python

machine-learning

mnist

tensorflow