tensorflow56 《TensorFlow技术解析与实战》06 神经网络的发展及其Tensorflow实现

# 《TensorFlow技术解析与实战》06 神经网络的发展及其TensorFlow实现
# win10 Tensorflow1.2.0-RC0 python3.5.3
# CUDA v8.0 cudnn-8.0-windows10-x64-v5.1
# filename:nntf06.01.py mnist的AlexNet实现
# 参考:
# https://github.com/aymericdamien/TensorFlow-Examples/blob/master/examples/3_NeuralNetworks/convolutional_network.py
# https://github.com/tensorflow/models/blob/master/tutorials/image/alexnet/alexnet_benchmark.py

import tensorflow as tf

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

# 定义网络的超参数
learing_rate = 0.001
training_iters = 200000
batch_size = 128
displayer_step = 10

# 定义网络的参数
n_input = 784    # 输入维度(img shape: 28x28)
n_classes = 10   # 标记维度(0-9 digits)
dropout = 0.75   # Dropout概率，输出的可能性

# 输入占位符
x = tf.placeholder(tf.float32, [None, n_input])
y = tf.placeholder(tf.float32, [None, n_classes])
keep_prob = tf.placeholder(tf.float32) # dropout

# 定义卷积操作
def conv2d(name, x, W, b, strides=1):
    x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')
    x = tf.nn.bias_add(x, b)
    return tf.nn.relu(x, name=name)  # 使用relu激活函数

# 定义池化操作
def maxpool2d(name, x, k=2):
    return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1], padding='SAME', name=name)

# 规范化操作
def norm(name, l_input, lsize=4):
    return tf.nn.lrn(l_input, lsize, bias=1.0, alpha=0.001/9.0, beta=0.75, name=name)

# 定义所有网络参数
weights = {
    'wc1': tf.Variable(tf.random_normal([11, 11, 1, 96])),
    'wc2': tf.Variable(tf.random_normal([5, 5, 96, 256])),
    'wc3': tf.Variable(tf.random_normal([3, 3, 256, 384])),
    'wc4': tf.Variable(tf.random_normal([3, 3, 384, 384])),
    'wc5': tf.Variable(tf.random_normal([3, 3, 384, 256])),
    'wd1': tf.Variable(tf.random_normal([4*4*256, 4096])),
    'wd2': tf.Variable(tf.random_normal([4096, 4096])),
    'out': tf.Variable(tf.random_normal([4096, 10]))
}
biases = {
    'bc1': tf.Variable(tf.random_normal([96])),
    'bc2': tf.Variable(tf.random_normal([256])),
    'bc3': tf.Variable(tf.random_normal([384])),
    'bc4': tf.Variable(tf.random_normal([384])),
    'bc5': tf.Variable(tf.random_normal([256])),
    'bd1': tf.Variable(tf.random_normal([4096])),
    'bd2': tf.Variable(tf.random_normal([4096])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}

# 定义网络
def alex_net(x, weights, biases, dropout):
    x = tf.reshape(x, shape=[-1, 28, 28, 1])

    conv1 = conv2d('conv1', x, weights['wc1'], biases['bc1'])
    pool1 = maxpool2d('pool1', conv1, k=2)
    norm1 = norm('norm1', pool1, lsize=4)

    conv2 = conv2d('conv2', conv1, weights['wc2'], biases['bc2'])
    pool2 = maxpool2d('pool2', conv2, k = 2)
    norm2 = norm('norm2', pool2, lsize=4)

    conv3 = conv2d('conv3', norm2, weights['wc3'], biases['bc3'])
    pool3 = maxpool2d('pool3', conv3, k = 2)
    norm3 = norm('norm3', pool3, lsize=4)

    conv4 = conv2d('conv4', norm3, weights['wc4'], biases['bc4'])
    conv5 = conv2d('conv5', norm3, weights['wc5'], biases['bc5'])
    pool5 = maxpool2d('pool5', conv5, k = 2)
    norm5 = norm('norm5', pool5, lsize=4)

    fc1 = tf.reshape(norm5, [-1, weights['wd1'].get_shape().as_list()[0]])
    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])
    fc1 = tf.nn.relu(fc1)
    fc1 = tf.nn.dropout(fc1, dropout)

    fc2 = tf.reshape(fc1, [-1, weights['wd2'].get_shape().as_list()[0]])
    fc2 = tf.add(tf.matmul(fc2, weights['wd2']), biases['bd2'])
    fc2 = tf.nn.relu(fc2)
    fc2 = tf.nn.dropout(fc2, dropout)

    out = tf.add(tf.matmul(fc2, weights['out']), biases['out'])
    return out

#构建预测模型
predict_model = alex_net(x, weights, biases, keep_prob)

# 定义损失函数和优化器
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=predict_model))
optimizer = tf.train.AdamOptimizer(learning_rate=learing_rate).minimize(cost)

# 评估函数
correct_pred = tf.equal(tf.argmax(predict_model, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# 训练模型和评估模型

# 初始化变量
init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)
    step = 1
    # 开始训练
    while step*batch_size < training_iters:
        batch_x, batch_y = mnist.train.next_batch(batch_size)
        sess.run(optimizer, feed_dict={x: batch_x, y: batch_y, keep_prob: dropout})
        if step % displayer_step == 0:
            # 计算损失值和准确度，输出
            loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_x, y: batch_y, keep_prob: 1.})
            print("Iter " + str(step*batch_size) + ", Minibatch Loss= " +
                  "{:.6f}".format(loss) + ", Training Accuracy= " +
                  "{:.5f}".format(acc))
        step += 1
    print("Optimizer Finished!")

    # 计算测试集的准确度
    print("Testing Accuracy: ", sess.run(accuracy, feed_dict={x: mnist.test.images[:256], y: mnist.test.labels[:256], keep_prob: 1.}))
'''
Iter 1280, Minibatch Loss= 460011.468750, Training Accuracy= 0.36719
Iter 2560, Minibatch Loss= 303076.562500, Training Accuracy= 0.62500
...
Iter 198400, Minibatch Loss= 4899.899414, Training Accuracy= 0.97656
Iter 199680, Minibatch Loss= 447.203613, Training Accuracy= 0.99219
Optimizer Finished!
Testing Accuracy:  0.992188
'''