MNIST數(shù)據(jù)集介紹

MNIST數(shù)據(jù)集中包含了各種各樣的手寫數(shù)字圖片，數(shù)據(jù)集的官網(wǎng)是：http://yann.lecun.com/exdb/mnist/index.html，我們可以從這里下載數(shù)據(jù)集。使用如下的代碼對數(shù)據(jù)集進行加載：

from tensorflow.examples.tutorials.mnist import input_datamnist = input_data.read_data_sets('MNIST_data', one_hot=True)

運行上述代碼會自動下載數(shù)據(jù)集并將文件解壓在MNIST_data文件夾下面。代碼中的one_hot=True，表示將樣本的標簽轉化為one_hot編碼。

MNIST數(shù)據(jù)集中的圖片是28*28的，每張圖被轉化為一個行向量，長度是28*28=784，每一個值代表一個像素點。數(shù)據(jù)集中共有60000張手寫數(shù)據(jù)圖片，其中55000張訓練數(shù)據(jù)，5000張測試數(shù)據(jù)。

在MNIST中，mnist.train.images是一個形狀為[55000, 784]的張量，其中的第一個維度是用來索引圖片，第二個維度圖片中的像素。MNIST數(shù)據(jù)集包含有三部分，訓練數(shù)據(jù)集，驗證數(shù)據(jù)集，測試數(shù)據(jù)集(mnist.validation)。

標簽是介于0-9之間的數(shù)字，用于描述圖片中的數(shù)字，轉化為one-hot向量即表示的數(shù)字對應的下標為1，其余的值為0。標簽的訓練數(shù)據(jù)是[55000,10]的數(shù)字矩陣。

下面定義了一個簡單的網(wǎng)絡對數(shù)據(jù)集進行訓練，代碼如下：

import tensorflow as tfimport numpy as npfrom tensorflow.examples.tutorials.mnist import input_dataimport matplotlib.pyplot as pltmnist = input_data.read_data_sets('MNIST_data', one_hot=True)tf.reset_default_graph()x = tf.placeholder(tf.float32, [None, 784])y = tf.placeholder(tf.float32, [None, 10])w = tf.Variable(tf.random_normal([784, 10]))b = tf.Variable(tf.zeros([10]))pred = tf.matmul(x, w) + bpred = tf.nn.softmax(pred)cost = tf.reduce_mean(-tf.reduce_sum(y * tf.log(pred), reduction_indices=1))learning_rate = 0.01optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)training_epochs = 25batch_size = 100display_step = 1save_path = 'model/'saver = tf.train.Saver()with tf.Session() as sess:  sess.run(tf.global_variables_initializer())  for epoch in range(training_epochs):    avg_cost = 0    total_batch = int(mnist.train.num_examples/batch_size)    for i in range(total_batch):      batch_xs, batch_ys = mnist.train.next_batch(batch_size)      _, c = sess.run([optimizer, cost], feed_dict={x:batch_xs, y:batch_ys})      avg_cost += c / total_batch    if (epoch + 1) % display_step == 0:      print('epoch= ', epoch+1, ' cost= ', avg_cost)  print('finished')  correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))  accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))  print('accuracy: ', accuracy.eval({x:mnist.test.images, y:mnist.test.labels}))  save = saver.save(sess, save_path=save_path+'mnist.cpkt')print(" starting 2nd session ...... ")with tf.Session() as sess:  sess.run(tf.global_variables_initializer())  saver.restore(sess, save_path=save_path+'mnist.cpkt')  correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))  accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))  print('accuracy: ', accuracy.eval({x: mnist.test.images, y: mnist.test.labels}))  output = tf.argmax(pred, 1)  batch_xs, batch_ys = mnist.test.next_batch(2)  outputval= sess.run([output], feed_dict={x:batch_xs, y:batch_ys})  print(outputval)  im = batch_xs[0]  im = im.reshape(-1, 28)  plt.imshow(im, cmap='gray')  plt.show()  im = batch_xs[1]  im = im.reshape(-1, 28)  plt.imshow(im, cmap='gray')  plt.show()