自编码器（AutoEncoder）

信息量本身是固有属性，我们如何通过合理的记忆方法或抽取方法去获取我们所在意的那部分有价值的信息？在信息冗余的这个时代，我们通过算法来简化信息量以便能合理分析解决问题------降维。

自编码器本身其实就是在做降维这件事，我们通过降维来获取对信息的偏好。自编码器的理解本身算比较简单，以图像处理为例。

当我们有一堆含有噪声的手写数字作为输入，然后乘上权重矩阵，得到编码矩阵C,这就是encoder了。此过程中得到的编码矩阵C往往小于之前的输入维度，本质就是降维的过程了。当然在深度学习的大环境下，deep layers 往往学得更好。那么我们如何来判定这个编码矩阵C能够表征我们的信息偏好呢？当然是以彼之道还施彼身了，之前encoder编码了信息，那么再做一下decoder解码，然后我们通过解码得到的输出与原始输入进行相似度计算，优化目标当然是两者越相似越好。decoder具体过程是通过编码矩阵C乘上权重矩阵，deep之后输出维度与输入维度相等。

以下是AutoEncoder代码实例，仅供参考。

from keras.layers import Input, Conv2D, MaxPooling2D, UpSampling2D
from keras.models import Model
from keras.datasets import mnist
from keras.callbacks import TensorBoard
from keras import backend as K
import numpy as np
import matplotlib.pyplot as plt
import pickle

input_img = Input(shape=(28, 28, 1))    # adapt this if using 'channels_first' image data format

x = Conv2D(16, (3, 3), activation='relu', padding='same')(input_img)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2, 2), padding='same')(x)

# at this point the representation is (4, 4, 8), i.e. 128-dimensional

x = Conv2D(8, (3, 3), activation='relu', padding='same')(encoded)
x = UpSampling2D((2, 2))(x)
x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
x = Conv2D(16, (3, 3), activation='relu')(x)
x = UpSampling2D((2, 2))(x)
decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)

autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')

# To train it, use the original MNIST digits with shape (samples, 3, 28, 28),
# and just normalize pixel values between 0 and 1

(x_train, _), (x_test, _) = mnist.load_data()

x_train = x_train.astype('float32') / 255.
x_test = x_test.astype('float32') / 255.
x_train = np.reshape(x_train, (len(x_train), 28, 28, 1))    # adapt this if using 'channels_first' image data format
x_test = np.reshape(x_test, (len(x_test), 28, 28, 1))       # adapt this if using 'channels_first' image data format

# open a terminal and start TensorBoard to read logs in the autoencoder subdirectory
# tensorboard --logdir=autoencoder

autoencoder.fit(x_train, x_train, epochs=50, batch_size=128, shuffle=True, validation_data=(x_test, x_test),
                callbacks=[TensorBoard(log_dir='conv_autoencoder')], verbose=2)

# take a look at the reconstructed digits
decoded_imgs = autoencoder.predict(x_test)

n = 10
plt.figure(figsize=(10, 4), dpi=100)
for i in range(n):
    # display original
    ax = plt.subplot(2, n, i + 1)
    plt.imshow(x_test[i].reshape(28, 28))
    plt.gray()
    ax.set_axis_off()

    # display reconstruction
    ax = plt.subplot(2, n, i + n + 1)
    plt.imshow(decoded_imgs[i].reshape(28, 28))
    plt.gray()
    ax.set_axis_off()

plt.show()

# take a look at the 128-dimensional encoded representation
# these representations are 8x4x4, so we reshape them to 4x32 in order to be able to display them as grayscale images

encoder = Model(input_img, encoded)
encoded_imgs = encoder.predict(x_test)

# save latent space features 128-d vector
pickle.dump(encoded_imgs, open('conv_autoe_features.pickle', 'wb'))

n = 10
plt.figure(figsize=(10, 4), dpi=100)
for i in range(n):
    ax = plt.subplot(1, n, i + 1)
    plt.imshow(encoded_imgs[i].reshape(4, 4 * 8).T)
    plt.gray()
    ax.set_axis_off()

plt.show()

 根据对自编码的描述，可以知道其具有广泛的应用场景。搜索查询（图、文相似搜索）、AI作画（decoder过程其实就是在产生新的图片数据）等