tensorflow实现AGN自编码器

1.载入tensorflow，MNIST数据集和常用库

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

注：导入数据集报错解决办法见：
http://blog.youkuaiyun.com/landcruiser007/article/details/79346982
2.定义参数初始化方法xavier initialization

def xavier_init(fan_in,fan_out,constant=1):
    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)

3.定义去噪自编码的class，代码较长，注意看注释

class AdditiveGaussianNoiseAutoencoder(object):
    def __init__(self,n_input,n_hidden,transfer_function=tf.nn.softplus,optimizer=tf.train.AdamOptimizer(),scale=0.1):
        #参数中transfer_function:隐含层激活函数，默认为softplus
        #参数中optimizer:优化器，默认为adam
        #scale:高斯噪声系数，默认0.1
        #n_input:输入变量数
        self.n_input=n_input
        #n_hidden:隐含层节点数
        self.n_hidden=n_hidden
        #transfer_function:隐含层激活函数，默认为softplus
        self.transfer=transfer_function

        self.scale=tf.placeholder(tf.float32)

        self.training_scale=scale
        #_initialize_weights参数初始化
        network_weights=self._initialize_weights()
        self.weights=network_weights
        #输入x
        self.x=tf.placeholder(tf.float32,[None,self.n_input])
        #建立能提取特征的隐含层，x加上噪声与隐含层的权重w1相乘，再加上隐含层的偏置b1，最后经过激活函数处理
        self.hidden=self.transfer(tf.add(tf.matmul(self.x+scale*tf.random_normal((n_input,)),
                                          self.weights['w1']),self.weights['b1']))
        #隐含层的输出乘w2+b2
        self.reconstruction=tf.add(tf.matmul(self.hidden,self.weights['w2']),self.weights['b2'])
        #自编码损失函数
        self.cost=0.5*tf.reduce_sum(tf.pow(tf.subtract(self.reconstruction,self.x),2.0))
        #优化器
        self.optimizer=optimizer.minimize(self.cost)
        init=tf.global_variables_initializer()
        #创建session
        self.sess=tf.Session()
        self.sess.run(init)
    #参数初始化函数
    def _initialize_weights(self):
        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
    #进行训练并返回损失
    def partial_fit(self,X):
        cost,opt=self.sess.run((self.cost,self.optimizer),feed_dict={self.x:X,self.scale:self.training_scale})
        return cost
    #只求损失，但不触发训练
    def calc_total_cost(self,X):
        return self.sess.run(self.cost,feed_dict={self.x:X,self.scale:self.training_scale})
    #返回自编码器隐含层的输出结果
    def transform(self,X):
        return self.sess.run(self.hidden,feed_dict={self.x:X,self.scale:self.training_scale})
    #重建
    def generate(self,hidden=None):
        if hidden is None:
            hidden=np.random.normal(size=self.weights["b1"])
        return self.sess.run(self.reconstruction,feed_dict={self.hidden:hidden})
    #复原
    def reconstruct(self,X):
        return self.sess.run(self.reconstruction,feed_dict={self.x:X,self.scale:self.training_scale})
    #获取隐含层权重w1
    def getWeights(self):
        return self.sess.run(self.weights['w1'])
    #获取隐含层偏置系数b1
    def getBiases(self):
        return self.sess.run(self.weights['b1'])

4.标准化处理
减去均值，除以标准差，得到均值为零，标准差为一的分布

def standard_scale(X_train,X_test):
    preprocessor=prep.StandardScaler().fit(X_train)
    X_train=preprocessor.transform(X_train)
    X_test=preprocessor.transform(X_test)
    return X_train,X_test

5.获取随机block数据
取0~（数据长度-batch size）的随机整数作为block的起始位置，取batch size个数据

def get_random_block_from_data(data,batch_size):
    start_index=np.random.randint(0,len(data)-batch_size)
    return data[start_index:(start_index+batch_size)]

6.对训练集，测试集进行标准化转换

X_train,X_test=standard_scale(mnist.train.images,mnist.test.images)

7.设置参数
最大训练轮数epochs：20
batch size：128
每训练一轮显示一次cost

n_samples=int(mnist.train.num_examples)
training_epochs=20
batch_size=128
display_step=1

8.创建AGN自编码器
第三步的class ：AdditiveGaussianNoiseAutoencoder
参数：
输入节点数：784
隐含层节点数：200
隐含层激活函数：softplus
优化器：adam，学习速率0.001
高斯噪声系数：0.01

autoencoder=AdditiveGaussianNoiseAutoencoder(n_input=784,
                                            n_hidden=200,
                                            transfer_function=tf.nn.softplus,
                                            optimizer=tf.train.AdamOptimizer(learning_rate=0.001),
                                            scale=0.01)

9.开始训练

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'%(epoch+1),"cost=","{:.9f}".format(avg_cost))

我得到的结果：

Epoch: 0001 cost= 7853.311381250
Epoch: 0002 cost= 7991.530777841
Epoch: 0003 cost= 7586.652882955
Epoch: 0004 cost= 7767.183337500
Epoch: 0005 cost= 7969.377174432
Epoch: 0006 cost= 7454.927330682
Epoch: 0007 cost= 7748.961891477
Epoch: 0008 cost= 8099.729705114
Epoch: 0009 cost= 7566.947763068
Epoch: 0010 cost= 7167.917351705
Epoch: 0011 cost= 7347.860416477
Epoch: 0012 cost= 7778.480061364
Epoch: 0013 cost= 7698.705884091
Epoch: 0014 cost= 7311.210797727
Epoch: 0015 cost= 7692.745509659
Epoch: 0016 cost= 8164.083282386
Epoch: 0017 cost= 7697.693646591
Epoch: 0018 cost= 7323.947401705
Epoch: 0019 cost= 8010.085572159
Epoch: 0020 cost= 8076.390184659

10.模型评测

print("total cost:"+str(autoencoder.calc_total_cost(X_test)))

我的结果是：

total cost:610581.0