深度學習筆記8：利用Tensorflow搭建神經網路
在筆記7中，筆者和大家一起入門了 Tensorflow 的基本語法，並舉了一些實際的例子進行了說明，終於告別了使用 numpy 手動搭建的日子。所以我們將繼續往下走，看看如何利用 Tensorflow 搭建神經網路模型。
儘管對於初學者而言使用 Tensorflow 看起來並不那麼習慣，需要各種步驟，但簡單來說， Tensorflow 搭建模型實際就是兩個過程：建立計算圖和執行計算圖。在 deeplearningai 課程中，NG和他的課程組給我們提供了 Signs Dataset （手勢）資料集，其中訓練集包括1080張64x64畫素的手勢圖片，並給定了 6 種標註，測試集包括120張64x64的手勢圖片，我們需要對訓練集構建神經網路模型然後對測試集給出預測。
先來簡單看一下資料集：
# Loading the datasetX_train_orig, Y_train_orig, X_test_orig, Y_test_orig, classes = load_dataset()# Flatten the training and test imagesX_train_flatten = X_train_orig.reshape(X_train_orig.shape[0], -1).T
X_test_flatten = X_test_orig.reshape(X_test_orig.shape[0], -1).T# Normalize image vectorsX_train = X_train_flatten/255.X_test = X_test_flatten/255.# Convert training and test labels to one hot matricesY_train = convert_to_one_hot(Y_train_orig, 6)
Y_test = convert_to_one_hot(Y_test_orig, 6)print ("number of training examples = " + str(X_train.shape[1]))print ("number of test examples = " + str(X_test.shape[1]))print ("X_train shape: " + str(X_train.shape))print ("Y_train shape: " + str(Y_train.shape))print ("X_test shape: " + str(X_test.shape))print ("Y_test shape: " + str(Y_test.shape))
下面就根據 NG 給定的找個資料集利用 Tensorflow 搭建神經網路模型。我們選擇構建一個包含 2 個隱層的神經網路，網路結構大致如下：
LINEAR -> RELU -> LINEAR -> RELU -> LINEAR -> SOFTMAX
正如我們之前利用 numpy 手動搭建一樣，搭建一個神經網路的主要步驟如下：
-定義網路結構
-初始化模型引數
-執行前向計算/計算當前損失/執行反向傳播/權值更新
建立 placeholder
根據 Tensorflow 的語法，我們首先建立輸入 X 和輸出 Y 的佔位符變數，這裡需要注意 shape 引數的設定。
def create_placeholders(n_x, n_y):
X = tf.placeholder(tf.float32, shape=(n_x, None), name='X')
Y = tf.placeholder(tf.float32, shape=(n_y, None), name='Y')
return X, Y
初始化模型引數
其次就是初始化神經網路的模型引數，三層網路包括六個引數，這裡我們採用 Xavier 初始化方法：
def initialize_parameters():
tf.set_random_seed(1)
W1 = tf.get_variable("W1", [25, 12288], initializer = tf.contrib.layers.xavier_initializer(seed = 1))
b1 = tf.get_variable("b1", [25, 1], initializer = tf.zeros_initializer())
b2 = tf.get_variable("b2", [12, 1], initializer = tf.zeros_initializer())
W2 = tf.get_variable("W2", [12, 25], initializer = tf.contrib.layers.xavier_initializer(seed = 1))
parameters = {"W1": W1,
W3 = tf.get_variable("W3", [6, 12], initializer = tf.contrib.layers.xavier_initializer(seed = 1)) b3 = tf.get_variable("b3", [6,1], initializer = tf.zeros_initializer())
"b1": b1,
"W2": W2,
"b2": b2,
"W3": W3,
"b3": b3}
return parameters
執行前向傳播
def forward_propagation(X, parameters): """
Implements the forward propagation for the model: LINEAR -> RELU -> LINEAR -> RELU -> LINEAR -> SOFTMAX
""" W1 = parameters['W1'] b1 = parameters['b1'] W2 = parameters['W2'] b2 = parameters['b2']
Z2 = tf.add(tf.matmul(W2, A1), b2)
W3 = parameters['W3'] b3 = parameters['b3'] Z1 = tf.add(tf.matmul(W1, X), b1) A1 = tf.nn.relu(Z1)
return Z3
A2 = tf.nn.relu(Z2)
Z3 = tf.add(tf.matmul(W3, A2), b3)
計算損失函式
在 Tensorflow 中損失函式的計算要比手動搭建時方便很多，一行程式碼即可搞定：
def compute_cost(Z3, Y):
logits = tf.transpose(Z3)
labels = tf.transpose(Y)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = logits, labels = labels))
return cost
程式碼整合：執行反向傳播和權值更新
跟計算損失函式類似， Tensorflow 中執行反向傳播的梯度優化非常簡便，兩行程式碼即可搞定，定義完整的神經網路模型如下：
def model(X_train, Y_train, X_test, Y_test, learning_rate = 0.0001,
num_epochs = 1500, minibatch_size = 32, print_cost = True):
(n_x, m) = X_train.shape
ops.reset_default_graph() tf.set_random_seed(1) seed = 3 n_y = Y_train.shape[0]
X, Y = create_placeholders(n_x, n_y) # Initialize parameters
costs = [] # Create Placeholders of shape (n_x, n_y)
parameters = initialize_parameters() # Forward propagation: Build the forward propagation in the tensorflow graph
Z3 = forward_propagation(X, parameters) # Cost function: Add cost function to tensorflow graph
optimizer = tf.train.GradientDescentOptimizer(learning_rate = learning_rate).minimize(cost) # Initialize all the variables
cost = compute_cost(Z3, Y) # Backpropagation: Define the tensorflow optimizer. Use an AdamOptimizer. init = tf.global_variables_initializer() # Start the session to compute the tensorflow graph
seed = seed + 1
with tf.Session() as sess: # Run the initialization sess.run(init) # Do the training loop for epoch in range(num_epochs): epoch_cost = 0. num_minibatches = int(m / minibatch_size)
minibatches = random_mini_batches(X_train, Y_train, minibatch_size, seed)
for minibatch in minibatches: # Select a minibatch
(minibatch_X, minibatch_Y) = minibatch
_ , minibatch_cost = sess.run([optimizer, cost], feed_dict={X: minibatch_X, Y: minibatch_Y})
epoch_cost += minibatch_cost / num_minibatches # Print the cost every epoch
if print_cost == True and epoch % 100 == 0:
print ("Cost after epoch %i: %f" % (epoch, epoch_cost))
if print_cost == True and epoch % 5 == 0:
costs.append(epoch_cost) # plot the cost
plt.ylabel('cost')
plt.plot(np.squeeze(costs)) plt.xlabel('iterations (per tens)')
plt.show() # lets save the parameters in a variable
plt.title("Learning rate =" + str(learning_rate))
parameters = sess.run(parameters)
print ("Parameters have been trained!") # Calculate the correct predictions
correct_prediction = tf.equal(tf.argmax(Z3), tf.argmax(Y)) # Calculate accuracy on the test set
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
print ("Train Accuracy:", accuracy.eval({X: X_train, Y: Y_train}))
print ("Test Accuracy:", accuracy.eval({X: X_test, Y: Y_test}))
return parameters
執行模型：
parameters = model(X_train, Y_train, X_test, Y_test)
根據模型的訓練誤差和測試誤差可以看到：模型整體效果雖然沒有達到最佳，但基本也能達到預測效果。
總結
● Tensorflow 語法中兩個基本的物件類是 Tensor 和 Operator.
● Tensorflow 執行計算的基本步驟為
● 建立計算圖（張量、變數和佔位符變數等）
● 建立會話
● 初始化會話
● 在計算圖中執行會話
可以看到的是，在 Tensorflow 中編寫神經網路要比我們手動搭建要方便的多，這也正是深度學習框架存在的意義之一。功能強大的深度學習框架能夠幫助我們快速的搭建起復雜的神經網路模型，在經歷了手動搭建神經網路的思維訓練過程之後，這對於我們來說就不再困難了。
原文釋出時間為：2018-09-9
本文作者：louwill