Tensorflow 实现 Softmax Regression 识别手写数字

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/",one_hot=True)
print(mnist.train.images.shape, mnist.train.labels.shape)
print(mnist.test.images.shape, mnist.test.labels.shape)
print(mnist.validation.images.shape, mnist.validation.labels.shape)

定义算法公式 y = softmax(Wx + b)

1
2
3

import tensorflow as tf
sess = tf.InteractiveSession()
x = tf.placeholder(tf.float32, [None, 784])

InteractiveSession 该方法将 sess注册为默认的session

1
2
3

W = tf.Variable(tf.zeros([784, 10]))
b = tf.Variable(tf.zeros([10]))
y = tf.nn.softmax(tf.matmul(x, W) + b)

matmul矩阵乘法

定义 loss，选定优化器

1 2	y_ = tf.placeholder(tf.float32, [None, 10]) cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))

tf.reduce_mean(x) 如果不指定第二个参数，那么就在所有的元素中取平均值
tf.reduce_mean(x, 0)

指定第二个参数为 0，则第一维的元素取平均值，即每一列求平均值

tf.reduce_sum

沿着reduction_indices所指定的维度求和。

1	train_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)

reduce_mean取均值
GradientDescentOptimizer梯度下降优化器

对数据进行迭代训练

tf.global_variables_initializer().run()
for i in range(1000):
    batch_xs, batch_ys = mnist.train.next_batch(100)
    train_step.run({x: batch_xs, y_: batch_ys})

评测准确率

1
2
3

correct_predicton = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_predicton, tf.float32))
print(accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))