TensorFlow入門示例教程

本部分的代碼目前都是基於GitHub大佬非常詳細的TensorFlow的教程上，首先給出鏈接：

https://github.com/aymericdamien/TensorFlow-Examples/

本人對其中部分代碼做了注釋和中文翻譯，會持續更新，目前包括：

　　1. 傳統多層神經網絡用語MNIST數據集分類（代碼講解，翻譯）

 

1. 傳統多層神經網絡用語MNIST數據集分類（代碼講解，翻譯）

 

""" Neural Network.

A 2-Hidden Layers Fully Connected Neural Network (a.k.a Multilayer Perceptron)
implementation with TensorFlow. This example is using the MNIST database
of handwritten digits (http://yann.lecun.com/exdb/mnist/).

Links:
    [MNIST Dataset](http://yann.lecun.com/exdb/mnist/).

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

from __future__ import print_function

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

# 導入tf
import tensorflow as tf

# Parameters
# 設定各種超參數
learning_rate = 0.1 # 學習率
num_steps = 500   # 訓練500次
batch_size = 128  # 每批次取128個樣本訓練
display_step = 100  # 每訓練100步顯示一次

# Network Parameters
# 設定網絡的超參數
n_hidden_1 = 256 # 1st layer number of neurons
n_hidden_2 = 256 # 2nd layer number of neurons
num_input = 784 # MNIST data input (img shape: 28*28)
num_classes = 10 # MNIST total classes (0-9 digits)

# tf Graph input
# tf圖的輸入，因為不知道到底輸入大小是多少，因此設定占位符
X = tf.placeholder("float", [None, num_input])
Y = tf.placeholder("float", [None, num_classes])

# Store layers weight & bias
# 初始化w和b
weights = {
    'h1': tf.Variable(tf.random_normal([num_input, n_hidden_1])),
    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),
    'out': tf.Variable(tf.random_normal([n_hidden_2, num_classes]))
}
biases = {
    'b1': tf.Variable(tf.random_normal([n_hidden_1])),
    'b2': tf.Variable(tf.random_normal([n_hidden_2])),
    'out': tf.Variable(tf.random_normal([num_classes]))
}


# Create model
# 創建模型
def neural_net(x):
    # Hidden fully connected layer with 256 neurons
    # 隱藏層1，全連接了256個神經元
    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])
    # Hidden fully connected layer with 256 neurons
    # 隱藏層2，全連接了256個神經元
    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])
    # Output fully connected layer with a neuron for each class
    # 最后作為輸出的全連接層，對每一分類連接一個神經元
    out_layer = tf.matmul(layer_2, weights['out']) + biases['out']
    return out_layer

# Construct model
# 開啟模型
# 輸入數據X，得到得分向量logits
logits = neural_net(X)
# 用softmax分類器將得分向量轉變成概率向量
prediction = tf.nn.softmax(logits)

# Define loss and optimizer
# 定義損失和優化器
# 交叉熵損失, 求均值得到---->loss_op
loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
    logits=logits, labels=Y))
# 優化器使用的是Adam算法優化器
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
# 最小化損失得到---->可以訓練的train_op
train_op = optimizer.minimize(loss_op)

# Evaluate model
# 評估模型
# tf.equal() 逐個元素進行判斷，如果相等就是True，不相等，就是False。
correct_pred = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))
# tf.cast() 數據類型轉換----> tf.reduce_mean() 再求均值
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# Initialize the variables (i.e. assign their default value)
# 初始化這些變量（作用比如說，給他們分配隨機默認值）
init = tf.global_variables_initializer()

# Start training
# 現在開始訓練啦！
with tf.Session() as sess:

    # Run the initializer
    # 運行初始化器
    sess.run(init)

    for step in range(1, num_steps+1):
        # 每批次128個訓練，取出這128個對應的data：x；標簽：y
        batch_x, batch_y = mnist.train.next_batch(batch_size)
        # Run optimization op (backprop)
        # train_op是優化器得到的可以訓練的op，通過反向傳播優化模型
        sess.run(train_op, feed_dict={X: batch_x, Y: batch_y})
        # 每100步打印一次訓練的成果
        if step % display_step == 0 or step == 1:
            # Calculate batch loss and accuracy
            # 計算每批次的是損失和准確度
            loss, acc = sess.run([loss_op, accuracy], feed_dict={X: batch_x,
                                                                 Y: batch_y})
            print("Step " + str(step) + ", Minibatch Loss= " + \
                  "{:.4f}".format(loss) + ", Training Accuracy= " + \
                  "{:.3f}".format(acc))

    print("Optimization Finished!")

    # Calculate accuracy for MNIST test images
    # 看看在測試集上，我們的模型表現如何
    print("Testing Accuracy:", \
        sess.run(accuracy, feed_dict={X: mnist.test.images,
                                      Y: mnist.test.labels}))