tensorflow实现Autoencoder自编码器生成Mnist手写数据集

本文内容只是方便自己下次学习，如有侵权，请联系我进行删除。内容主要是《Tensorflow实战》中第四章实现自编码器

参考博客：https://blog.csdn.net/louishao/article/details/76218083

参考博客：https://blog.csdn.net/marsjhao/article/details/68950697

一、在tensorflow下加载Mnist

import numpy as np
import sklearn.preprocessing as prep
import tensorflow as tf
import matplotlib.pyplot as plt 
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets( '../MNIST_data', one_hot = True )

二、权重初始大小设置

def xavier_init(fan_in,fan_out,constant=1):#DNN权重初始化太小，每层传递时候缩小难以产生作用。太大会导致发散和失效
    low  = -constant * np.sqrt( 6.0 / ( fan_in + fan_out ) )
    high =  constant * np.sqrt( 6.0 / ( fan_in + fan_out ) )
    return tf.random_uniform((fan_in, fan_out), minval=low, maxval=high, dtype=tf.float32 )

三 、创建自编码器类

class AdditiveGaussianNoiseAutoencoder(object):
    def __init__(self, n_input, n_hidden, transfer_function=tf.nn.softplus, 
                 optimizer = tf.train.AdamOptimizer(), scale=0.1 ):
        self.n_input = n_input
        self.n_hidden = n_hidden
        self.transfer = transfer_function
        self.scale = tf.placeholder( tf.float32 )
        self.training_scale = scale
        network_weights = self._initialize_weights()
        self.weights = network_weights
        
        self.x = tf.placeholder( tf.float32, [None, self.n_input] )
        self.hidden = self.transfer( tf.add( tf.matmul(self.x + scale * tf.random_normal(( n_input, ) ), self.weights['w1'] ), self.weights['b1'] ))
        self.reconstruction = tf.add( tf.matmul( self.hidden, self.weights['w2'] ), self.weights['b2'] )
        
        self.cost = 0.5 * tf.reduce_sum( tf.pow( tf.subtract( self.reconstruction, self.x ), 2 ) )
        self.optimizer = optimizer.minimize( self.cost )
        init = tf.global_variables_initializer()
        self.sess = tf.Session()
        self.sess.run( init )
        
    def _initialize_weights(self):#初始化所有权重
        all_weights=dict()
        all_weights['w1']=tf.Variable(xavier_init(self.n_input,self.n_hidden))
        all_weights['b1']=tf.Variable(tf.zeros([self.n_hidden],dtype=tf.float32))
        all_weights['w2']=tf.Variable(tf.zeros([self.n_hidden,self.n_input],dtype=tf.float32))
        all_weights['b2']=tf.Variable(tf.zeros([self.n_input],dtype=tf.float32))
    
        return all_weights

    def partial_fit(self, X):#执行一步训练
        cost, opt = self.sess.run( (self.cost, self.optimizer), 
                                feed_dict = { self.x : X, self.scale : self.training_scale } )
        return cost
    def calc_total_cost( self, X ):#求cost
        return self.sess.run( self.cost, feed_dict = { self.x : X, self.scale : self.training_scale } )
    
    def transform( self, X ):#求隐藏
        return self.sess.run( self.hidden, feed_dict = { self.x : X, self.scale : self.training_scale } )

    def generate( self, hidden = None ):#求最后的生成层
        if hidden == None:
            hidden = np.random.normal( size = self.weights['b1'] )
        return self.sess.run( self.reconstruction, feed_dict = { self.hidden : hidden } )

    def reconstructData( self, X ):#整体运行一遍复原过程
        return self.sess.run( self.reconstruction, feed_dict = { self.x : X, self.scale : self.training_scale } )

    def getWeights( self ):#获取参数w1
        return self.sess.run( self.weights['w1'] )

    def getBiases( self ):#获取参数w2
        return self.sess.run( self.weights['b1'] )
    # 可视化对比原输入图像和加入噪声后的图像
    def plot_noiseimg(self,img,show_comp=True):
        self.noise = self.sess.run(tf.random_normal((self.n_input,)))
        noiseimg = img + self.training_scale*self.noise
        plot_image(noiseimg)
        if show_comp:
            plt.subplot(121)
            plt.imshow(img.reshape((28, 28)), interpolation='nearest', cmap='binary')
            plt.subplot(122)
            plot_image(noiseimg)

四、标准化数据

def standard_scale( X_train, X_test ):#数据标准化
    preprocessor = prep.StandardScaler().fit( X_train )
    X_train = preprocessor.transform( X_train )
    X_test  = preprocessor.transform( X_test )
    return X_train, X_test
X_train, X_test  =standard_scale( mnist.train.images, mnist.test.images )

五、为训练分block

def get_random_block_from_data( data, batch_size ):#获取随机block,随机数作为起始位置，不放回抽样
    start_index = np.random.randint( 0, len(data) - batch_size )
    return data[ start_index : (start_index+batch_size)  ]

六、训练

n_samples = int( mnist.train.num_examples )#总训练样本数
training_epochs = 20
batch_size = 128
display_step = 1#每隔1论显示cost

autoencoder = AdditiveGaussianNoiseAutoencoder( n_input = 784,n_hidden = 200, 
                       
                                               transfer_function = tf.nn.softplus,
                                               optimizer = tf.train.AdamOptimizer(learning_rate = 0.0001),
                                               scale = 0.01 )#scale为噪声系数，这是创建一个AGN自编码器

for epoch in range( training_epochs ):
    avg_cost = 0
    total_batch = int( n_samples / batch_size )
    for i in range( total_batch ):
        batch_xs = get_random_block_from_data( X_train, batch_size )

        cost = autoencoder.partial_fit( batch_xs )
        avg_cost = cost / n_samples * batch_size

    if epoch % display_step == 0:
        print( "epoch : %04d, cost = %.9f" % ( epoch+1, avg_cost ) )

七、查看cost

print( "Total cost : ",  str( autoencoder.calc_total_cost(X_test)))

八、查看生成图片

encode_decode=autoencoder.reconstructData(X_test)#注意reconstructData，不能与reconstruction一样,书上是reconstruction
f, a = plt.subplots(2, 10, figsize=(10, 2))  
examples_to_show=10
for i in range(examples_to_show):  
    a[0][i].imshow(np.reshape(mnist.test.images[i], (28, 28)))  
    a[1][i].imshow(np.reshape(encode_decode[i], (28, 28)))   
plt.show()