在筆記7中，筆者和大家一起入門了 Tensorflow的基本語法，并舉了一些實際的例子進行了說明，終于告別了使用numpy手動搭建的日子。所以我們將繼續(xù)往下走，看看如何利用 Tensorflow搭建神經(jīng)網(wǎng)絡(luò)模型。

盡管對于初學(xué)者而言使用Tensorflow看起來并不那么習(xí)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：創(chuàng)建計算圖和執(zhí)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）數(shù)據(jù)集，其中訓(xùn)練集包括1080張64x64像素的手勢圖片，并給定了 6 種標(biāo)注，測試集包括120張64x64的手勢圖片，我們需要對訓(xùn)練集構(gòu)建神經(jīng)網(wǎng)絡(luò)模型然后對測試集給出預(yù)測。

先來簡單看一下數(shù)據(jù)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(jù) NG 給定的找個數(shù)據(jù)集利用Tensorflow搭建神經(jīng)網(wǎng)絡(luò)模型。我們選擇構(gòu)建一個包含 2 個隱層的神經(jīng)網(wǎng)絡(luò)，網(wǎng)絡(luò)結(jié)構(gòu)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(jīng)網(wǎng)絡(luò)的主要步驟如下：-定義網(wǎng)絡(luò)結(jié)構(gòu)-初始化模型參數(shù)-執(zhí)行前向計算/計算當(dāng)前損失/執(zhí)行反向傳播/權(quán)值更新

創(chuàng)建 placeholder

根據(jù)Tensorflow的語法，我們首先創(chuàng)建輸入X和輸出Y的占位符變量，這里需要注意shape參數(shù)的設(shè)置。

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

初始化模型參數(shù)

其次就是初始化神經(jīng)網(wǎng)絡(luò)的模型參數(shù)，三層網(wǎng)絡(luò)包括六個參數(shù)，這里我們采用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()) W2 = tf.get_variable("W2", [12, 25], initializer = tf.contrib.layers.xavier_initializer(seed = 1)) b2 = tf.get_variable("b2", [12, 1], initializer = tf.zeros_initializer()) 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()) parameters = {"W1": W1, "b1": b1, "W2": W2, "b2": b2, "W3": W3, "b3": b3} return parameters

執(zhí)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'] W3 = parameters['W3'] b3 = parameters['b3'] Z1 = tf.add(tf.matmul(W1, X), b1) A1 = tf.nn.relu(Z1) Z2 = tf.add(tf.matmul(W2, A1), b2) A2 = tf.nn.relu(Z2) Z3 = tf.add(tf.matmul(W3, A2), b3) return Z3

計算損失函數(shù)

在Tensorflow中損失函數(shù)的計算要比手動搭建時方便很多，一行代碼即可搞定：

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

代碼整合：執(zhí)行反向傳播和權(quán)值更新

跟計算損失函數(shù)類似，Tensorflow中執(zhí)行反向傳播的梯度優(yōu)化非常簡便，兩行代碼即可搞定，定義完整的神經(jīng)網(wǎng)絡(luò)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): ops.reset_default_graph() tf.set_random_seed(1) seed = 3 (n_x, m) = X_train.shape n_y = Y_train.shape[0] costs = [] # Create Placeholders of shape (n_x, n_y) X, Y = create_placeholders(n_x, n_y) # Initialize parameters 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 cost = compute_cost(Z3, Y) # Backpropagation: Define the tensorflow optimizer. Use an AdamOptimizer. optimizer = tf.train.GradientDescentOptimizer(learning_rate = learning_rate).minimize(cost) # Initialize all the variables init = tf.global_variables_initializer() # Start the session to compute the tensorflow graph 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) seed = seed + 1 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.plot(np.squeeze(costs)) plt.ylabel('cost') plt.xlabel('iterations (per tens)') plt.title("Learning rate =" + str(learning_rate)) plt.show() # lets save the parameters in a variable 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(zhí)行模型：

parameters = model(X_train, Y_train, X_test, Y_test)

根據(jù)模型的訓(xùn)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(yù)測效果。

總結(jié)

Tensorflow語法中兩個基本的對象類是 Tensor 和 Operator.

Tensorflow執(zhí)行計算的基本步驟為

創(chuàng)建計算圖（張量、變量和占位符變量等）

創(chuàng)建會話

初始化會話

在計算圖中執(zhí)行會話

可以看到的是，在 Tensorflow 中編寫神經(jīng)網(wǎng)絡(luò)要比我們手動搭建要方便的多，這也正是深度學(xué)習(xí)框架存在的意義之一。功能強大的深度學(xué)習(xí)框架能夠幫助我們快速的搭建起復(fù)雜的神經(jīng)網(wǎng)絡(luò)模型，在經(jīng)歷了手動搭建神經(jīng)網(wǎng)絡(luò)的思維訓(xùn)練過程之后，這對于我們來說就不再困難了。