使用tf.layers.batch_normalization()需要三步：

• 在卷积层将激活函数设置为None。
• 使用batch_normalization。

• 使用激活函数激活。
  需要特别注意的是：在训练时，需要将第二个参数training = True。在测试时，将training = False。同时，在降低loss时候时候：

    update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
    with tf.control_dependencies(update_ops):
        train_op = tf.train.AdamOptimizer(1e-3).minimize(loss) #使用AdamOptimizer优化器将损失函数降到最低

  这里要注意如果不添加update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)会导致训练模式下的预测正确率很好，但是在预测模式测试集中最后计算的正确率非常低。

AlexNet的Tensorflow实现

环境

• python3.6
• tensorflow 0.10.0

建议使用anoconda安装，可以节省不少时间和方便不少

加载Cifar10的数据集
数据集使用cifar10，需要自行下载

def load_data(filename):
    """read data from data file"""
    with open(filename,'rb') as f:
        data = pickle.load(f,encoding='latin1')
        return data['data'],data['labels']
class CifarData:
    def __init__(self, filenames, need_shuffle):
        all_data = []
        all_labels = []
        for filename in filenames:
            data, labels = load_data(filename)
            all_data.append(data)
            all_labels.append(labels)
        self._data = np.vstack(all_data)
        self._data = self._data / 127.5 - 1
        self._labels = np.hstack(all_labels)
        self._num_examples = self._data.shape[0]
        self._need_shuffle = need_shuffle
        self._indicator = 0
        if self._need_shuffle:
            self._shuffle_data()
    def _shuffle_data(self):
        p = np.random.permutation(self._num_examples)
        self._data = self._data[p]
        self._labels = self._labels[p]

    def next_batch(self, batch_size):
        """return batch_size examples as a batch"""
        end_indicator = self._indicator + batch_size
        if end_indicator > self._num_examples:
            if self._need_shuffle:
                self._shuffle_data()
                self._indicator = 0
                end_indicator = batch_size
            else:
                raise Exception("Have no more examples")
        if end_indicator > self._num_examples:
            raise Exception ("Batch size is larger than all examples")
        batch_data = self._data[self._indicator:end_indicator]
        batch_lebel = self._labels[self._indicator:end_indicator]
        self._indicator = end_indicator
        return batch_data, batch_lebel

网络部分代码,这里使用了一下手动实现batch_normalization，也就是批归一化的流程

def batch_normal(xs, out_size):
    axis = list(range(len(xs.get_shape()) - 1))
    n_mean, n_var = tf.nn.moments(xs, axes=axis)
    scale = tf.Variable(tf.ones([out_size]))
    shift = tf.Variable(tf.zeros([out_size]))
    epsilon = 0.001
    ema = tf.train.ExponentialMovingAverage(decay=0.9)
    
    def mean_var_with_update():
        ema_apply_op = ema.apply([n_mean, n_var])
        with tf.control_dependencies([ema_apply_op]):
            return tf.identity(n_mean), tf.identity(n_var)
        
    mean, var = mean_var_with_update()
        
    bn = tf.nn.batch_normalization(xs, mean, var, shift, scale, epsilon)
    return bn

下面是正常的网络流程

这里放一张网络结构图，我看目前搜索引擎很多都只有图但是没有这个最早期在CNN网络之前的代码，所以自己尝试写了一下，也算学习一下

AlexNet网络结构：

由于早期是显卡瓶颈导致的需要两张显卡做运算之后交叉数据，现在一张显卡就可以胜任这个工作了，所以和图上的结构有些许区别。

train_filenames = [os.path.join(CIFAR_DIR, 'data_batch_%d' % i) for i in range(1,6)]
test_filenames = [os.path.join(CIFAR_DIR, 'test_batch')]

train_data = CifarData(train_filenames, True)

x = tf.placeholder(tf.float32,[None,3072])
# [None], rg [0,6,5,3]
y = tf.placeholder(tf.int64,[None])
is_training = tf.placeholder(tf.bool,[])
x_image = tf.reshape(x,[-1,3,32,32])
x_image = tf.transpose(x_image, perm=[0, 2, 3, 1])

# 神经元图 feature_map, 输出图像
# 32*32
conv1 = tf.layers.conv2d(x_image,
                        48,#output channel number
                        (3,3),#kenel size
                        padding = 'same',
                        activation=tf.nn.relu,
                        name='conv1')

lrn1 = tf.nn.lrn(conv1, depth_radius=4, bias=1.0, alpha=0.001 / 9.0, beta=0.75)

# 16*16
pooling1 = tf.layers.max_pooling2d(lrn1,
                                    (2,2),
                                    (2,2),
                                    name='pool1')
# 16*16
conv2 = tf.layers.conv2d(pooling1,
                        96,#output channel number
                        (3,3),#kenel size
                        padding = 'same',
                        activation=None,
                        name='conv2')
# batch_normalization

bn = tf.layers.batch_normalization(conv2, training=is_training )

conv2 = tf.nn.relu(bn)

# 取消使用lrn层改成bn
# lrn2 = tf.nn.lrn(conv2, depth_radius=4, bias=1.0, alpha=0.001 / 9.0, beta=0.75)

# 8*8
pooling2 = tf.layers.max_pooling2d(conv2,
                                    (2,2),
                                    (2,2),
                                    name='pool2')
# 8*8
conv3 = tf.layers.conv2d(pooling2,
                        192,#output channel number
                        (3,3),#kenel size
                        padding = 'same',
                        activation=tf.nn.relu,
                        name='conv3')
# 8*8
conv4 = tf.layers.conv2d(conv3,
                        192,#output channel number
                        (3,3),#kenel size
                        padding = 'same',
                        activation=tf.nn.relu,
                        name='conv4')
# 8*8
conv5 = tf.layers.conv2d(conv4,
                        96,#output channel number
                        (3,3),#kenel size
                        padding = 'same',
                        activation=tf.nn.relu,
                        name='conv5')

bn = tf.layers.batch_normalization(conv5, training=is_training )

conv5 = tf.nn.relu(bn)

# 4*4
pooling5 = tf.layers.max_pooling2d(conv5,
                                    (2,2),
                                    (2,2),
                                    name='pool2')

# [None, 1024]
flatten = tf.layers.flatten(pooling5)
y_1 = tf.layers.dense(flatten, 1024)
bn6 = batch_normal(y_1, 1024)
fc1 = tf.nn.relu(bn6)
# [None, 1024]
y_2 = tf.layers.dense(fc1, 1024)
bn7 = batch_normal(y_2, 1024)
fc2 = tf.nn.relu(bn7)
# [None, 10]
y_ = tf.layers.dense(fc2, 10)

# 交叉熵损失函数
# y_->softmax
# y -> one_hot
# loss = ylogy_
loss = tf.losses.sparse_softmax_cross_entropy(labels=y, logits=y_)

# bool
predict = tf.argmax(y_, 1)
correct_prediction = tf.equal(predict, y)
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float64))

with tf.name_scope('train_op'):
    update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
    with tf.control_dependencies(update_ops):
        train_op = tf.train.AdamOptimizer(1e-3).minimize(loss)

init = tf.global_variables_initializer()
batch_size = 20
test_steps = 100
train_steps = 10000
with tf.Session() as sess:
    sess.run(init)
    for i in range(train_steps):
        batch_data, batch_labels = train_data.next_batch(batch_size)
        loss_val, accu_val, _ = sess.run([loss, accuracy, train_op], feed_dict={x: batch_data, y: batch_labels, is_training:True})
        if (i+1) % 500 == 0:
            print('[Train] Step: %d, loss: %4.5f, acc:%4.5f' % (i+1, loss_val, accu_val))
        if (i+1) % 5000 == 0:
            test_data = CifarData(test_filenames, False)
            all_test_acc_val = []
            for j in range(test_steps):
                test_bach_data, test_batch_labels = test_data.next_batch(batch_size)
                test_acc_val = sess.run([accuracy], feed_dict={x: test_bach_data, y: test_batch_labels, is_training:False})
                all_test_acc_val.append(test_acc_val)
            test_acc = np.mean(all_test_acc_val)
            print("[Test] Step: %d, acc:%4.5f" %(i+1, test_acc))

下面是在运行了10000次之后的测试结果：

可以看出测试集正确率可以在73.7%左右。算是比较高的了在AlexNet这个网络下面。