TensorFlow2.0 代码实战专栏（八）：双向循环神经网络示例

from __future__ import print_function

import tensorflow as tf
from tensorflow.contrib import rnn
import numpy as np

# 导入MNIST数据
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)

Extracting /tmp/data/train-images-idx3-ubyte.gz

Extracting /tmp/data/train-labels-idx1-ubyte.gz

Extracting /tmp/data/t10k-images-idx3-ubyte.gz

Extracting /tmp/data/t10k-labels-idx1-ubyte.gz

# 训练参数
learning_rate = 0.001
training_steps = 10000
batch_size = 128
display_step = 200

# 网络参数
num_input = 28 # MNIST数据输入 (图像形状: 28*28)
timesteps = 28 # 时间步长
num_hidden = 128 # 隐藏层特征数
num_classes = 10 # 所有类别（数字 0-9）

# tf 图输入
X = tf.placeholder("float", [None, timesteps, num_input])
Y = tf.placeholder("float", [None, num_classes])

# 定义权重
weights = {
    #隐含层权重值=> 2*n_hidden，因为前向+后向单元
    'out': tf.Variable(tf.random_normal([2*num_hidden, num_classes]))
}
biases = {
    'out': tf.Variable(tf.random_normal([num_classes]))
}

def BiRNN(x, weights, biases):

    # 准备数据形状以符合rnn函数要求
    # 当前数据输入形状: (batch_size, timesteps, n_input)
    # 要求的形状: 形状为'timesteps'个张量的列表 (batch_size, num_input)

    # 分解得到形状为'timesteps'个张量的列表形状为'timesteps'个张量的列表  
    x = tf.unstack(x, timesteps, 1)

    # 使用tensorflow定义lstm单元
    # 前向单元
    lstm_fw_cell = rnn.BasicLSTMCell(num_hidden, forget_bias=1.0)
    # 后向单元
    lstm_bw_cell = rnn.BasicLSTMCell(num_hidden, forget_bias=1.0)

    # 得到lstm单元输出
    try:
        outputs, _, _ = rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x,
                                              dtype=tf.float32)
    except Exception: # 旧的TensorFlow版本只返回输出，而不是状态
        outputs = rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x,
                                        dtype=tf.float32)

    # 线性激活，使用rnn内部循环最后的输出
    return tf.matmul(outputs[-1], weights['out']) + biases['out']

logits = BiRNN(X, weights, biases)
prediction = tf.nn.softmax(logits)

# 定义损失和优化器
loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
    logits=logits, labels=Y))
optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss_op)

# 评估模型(测试时，禁用dropout)
correct_pred = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# 初始化变量(即分配它们的默认值)
init = tf.global_variables_initializer()

# 开始训练
with tf.Session() as sess:

    # 运行初始化
    sess.run(init)

    for step in range(1, training_steps+1):
        batch_x, batch_y = mnist.train.next_batch(batch_size)
        # 改变数据形状以获得28个元素的28个序列
        batch_x = batch_x.reshape((batch_size, timesteps, num_input))
        # 运行优化操作 (反向传播)
        sess.run(train_op, feed_dict={X: batch_x, Y: batch_y})
        if step % display_step == 0 or step == 1:
            # 计算批损失和准确率
            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!")

    # 计算128个mnist测试图像的准确率
    test_len = 128
    test_data = mnist.test.images[:test_len].reshape((-1, timesteps, num_input))
    test_label = mnist.test.labels[:test_len]
    print("Testing Accuracy:", \
        sess.run(accuracy, feed_dict={X: test_data, Y: test_label}))

Step 1, Minibatch Loss= 2.6218, Training Accuracy= 0.086

Step 200, Minibatch Loss= 2.1900, Training Accuracy= 0.211

Step 400, Minibatch Loss= 2.0144, Training Accuracy= 0.375

Step 600, Minibatch Loss= 1.8729, Training Accuracy= 0.445

Step 800, Minibatch Loss= 1.8000, Training Accuracy= 0.469

Step 1000, Minibatch Loss= 1.7244, Training Accuracy= 0.453

Step 1200, Minibatch Loss= 1.5657, Training Accuracy= 0.523

Step 1400, Minibatch Loss= 1.5473, Training Accuracy= 0.547

Step 1600, Minibatch Loss= 1.5288, Training Accuracy= 0.500

Step 1800, Minibatch Loss= 1.4203, Training Accuracy= 0.555

Step 2000, Minibatch Loss= 1.2525, Training Accuracy= 0.641

Step 2200, Minibatch Loss= 1.2696, Training Accuracy= 0.594

Step 2400, Minibatch Loss= 1.2000, Training Accuracy= 0.664

Step 2600, Minibatch Loss= 1.1017, Training Accuracy= 0.625

Step 2800, Minibatch Loss= 1.2656, Training Accuracy= 0.578

Step 3000, Minibatch Loss= 1.0830, Training Accuracy= 0.656

Step 3200, Minibatch Loss= 1.1522, Training Accuracy= 0.633

Step 3400, Minibatch Loss= 0.9484, Training Accuracy= 0.680

Step 3600, Minibatch Loss= 1.0470, Training Accuracy= 0.641

Step 3800, Minibatch Loss= 1.0609, Training Accuracy= 0.586

Step 4000, Minibatch Loss= 1.1853, Training Accuracy= 0.648

Step 4200, Minibatch Loss= 0.9438, Training Accuracy= 0.750

Step 4400, Minibatch Loss= 0.7986, Training Accuracy= 0.766

Step 4600, Minibatch Loss= 0.8070, Training Accuracy= 0.750

Step 4800, Minibatch Loss= 0.8382, Training Accuracy= 0.734

Step 5000, Minibatch Loss= 0.7397, Training Accuracy= 0.766

Step 5200, Minibatch Loss= 0.7870, Training Accuracy= 0.727

Step 5400, Minibatch Loss= 0.6380, Training Accuracy= 0.828

Step 5600, Minibatch Loss= 0.7975, Training Accuracy= 0.719

Step 5800, Minibatch Loss= 0.7934, Training Accuracy= 0.766

Step 6000, Minibatch Loss= 0.6628, Training Accuracy= 0.805

Step 6200, Minibatch Loss= 0.7958, Training Accuracy= 0.672

Step 6400, Minibatch Loss= 0.6582, Training Accuracy= 0.773

Step 6600, Minibatch Loss= 0.5908, Training Accuracy= 0.812

Step 6800, Minibatch Loss= 0.6182, Training Accuracy= 0.820

Step 7000, Minibatch Loss= 0.5513, Training Accuracy= 0.812

Step 7200, Minibatch Loss= 0.6683, Training Accuracy= 0.789

Step 7400, Minibatch Loss= 0.5337, Training Accuracy= 0.828

Step 7600, Minibatch Loss= 0.6428, Training Accuracy= 0.805

Step 7800, Minibatch Loss= 0.6708, Training Accuracy= 0.797

Step 8000, Minibatch Loss= 0.4664, Training Accuracy= 0.852

Step 8200, Minibatch Loss= 0.4249, Training Accuracy= 0.859

Step 8400, Minibatch Loss= 0.7723, Training Accuracy= 0.773

Step 8600, Minibatch Loss= 0.4706, Training Accuracy= 0.859

Step 8800, Minibatch Loss= 0.4800, Training Accuracy= 0.867

Step 9000, Minibatch Loss= 0.4636, Training Accuracy= 0.891

Step 9200, Minibatch Loss= 0.5734, Training Accuracy= 0.828

Step 9400, Minibatch Loss= 0.5548, Training Accuracy= 0.875

Step 9600, Minibatch Loss= 0.3575, Training Accuracy= 0.922

Step 9800, Minibatch Loss= 0.4566, Training Accuracy= 0.844

Step 10000, Minibatch Loss= 0.5125, Training Accuracy= 0.844

Optimization Finished!

Testing Accuracy: 0.890625