#來源於莫煩tensorflow視頻中學習
# -*- coding: utf-8 -*-
"""autoencoder mnist
can running
autoencoder自定義實現，未直接調用函數，顯示autoencoder結果與原來真實輸入數據的對比圖
"""
#特色：可視化 通過encoder最後一層神經元數目爲2，將數據降維到2維，進行畫點plt.scatter可視化
#劃分的不咋開
#import packages
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from tensorflow.examples.tutorials.mnist import input_data
#load data
mnist = input_data.read_data_sets('MNIST_data/', one_hot=False)
#define hyperparameter
learning_rate = 0.0001
traning_epochs = 20
batch_size = 256
display_step = 1
examples_to_show = 10
# 訓練training_epochs=5個epochs,每個epoch裏面有batch_size筆data
# examples_to_show = 10用於測試 autoencoder結果與真實data的對比，畫出對比圖
n_input = 784 # MNIST data input shape(28*28)
#input variable shape
X = tf.placeholder("float", [None, n_input])
#define hidden layers
n_hidden_1 = 256
n_hidden_2 = 64
n_hidden_3 = 10
n_hidden_4 = 2
# define encoder and decoder variables
weights={ #encoder 與 decoder 是對稱的，包括activation function
'encoder_h1':tf.Variable(tf.random_normal([n_input, n_hidden_1])),
'encoder_h2':tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),
'encoder_h3':tf.Variable(tf.random_normal([n_hidden_2, n_hidden_3])),
'encoder_h4':tf.Variable(tf.random_normal([n_hidden_3, n_hidden_4])),
'decoder_h1':tf.Variable(tf.random_normal([n_hidden_4, n_hidden_3])),
'decoder_h2':tf.Variable(tf.random_normal([n_hidden_3, n_hidden_2])),
'decoder_h3':tf.Variable(tf.random_normal([n_hidden_2, n_hidden_1])),
'decoder_h4':tf.Variable(tf.random_normal([n_hidden_1, n_input]))
}
biases={
'encoder_b1':tf.Variable(tf.random_normal([n_hidden_1,])),
'encoder_b2':tf.Variable(tf.random_normal([n_hidden_2,])),
'encoder_b3':tf.Variable(tf.random_normal([n_hidden_3,])),
'encoder_b4':tf.Variable(tf.random_normal([n_hidden_4,])),
'decoder_b1':tf.Variable(tf.random_normal([n_hidden_3,])),
'decoder_b2':tf.Variable(tf.random_normal([n_hidden_2,])),
'decoder_b3':tf.Variable(tf.random_normal([n_hidden_1,])),
'decoder_b4':tf.Variable(tf.random_normal([n_input,]))
}
def encoder(x):# y = w * x + b
layer_1 = tf.nn.sigmoid(tf.add(tf.matmul(x, weights['encoder_h1']), biases['encoder_b1']))
layer_2 = tf.nn.sigmoid(tf.add(tf.matmul(layer_1, weights['encoder_h2']), biases['encoder_b2']))
layer_3 = tf.nn.sigmoid(tf.add(tf.matmul(layer_2, weights['encoder_h3']), biases['encoder_b3']))
# linear activation function
layer_4 = tf.add(tf.matmul(layer_3, weights['encoder_h4']), biases['encoder_b4'])
return layer_4
def decoder(x): # encoder outputs as the inputs of decoder
layer_1 = tf.nn.sigmoid(tf.add(tf.matmul(x, weights['decoder_h1']), biases['decoder_b1']))
# tf.nn.sigmoid將數據範圍歸一化到max(x)=1，min(x)=0
layer_2 = tf.nn.sigmoid(tf.add(tf.matmul(layer_1, weights['decoder_h2']), biases['decoder_b2']))
layer_3 = tf.nn.sigmoid(tf.add(tf.matmul(layer_2, weights['decoder_h3']), biases['decoder_b3']))
# tf.nn.sigmoid將數據範圍歸一化到max(x)=1，min(x)=0
layer_4 = tf.nn.sigmoid(tf.add(tf.matmul(layer_3, weights['decoder_h4']), biases['decoder_b4']))
return layer_4
#operation return results
encoder_op = encoder(X)
decoder_op = decoder(encoder_op)
# unsuperivised constrast
y_pred = decoder_op
y_true = X
#define cost mean square error
cost = tf.reduce_mean(tf.pow(y_true - y_pred, 2))
#define train operation
optimizer = tf.train.AdamOptimizer(learning_rate).minimize(cost)
init = tf.initialize_all_variables()
# mnist.train.images 用於訓練
with tf.Session() as sess:
sess.run(init)
# 數據可以劃分成多少個batch,有多少個batch，可以進行多少輪訓練,每個batch裏面有batch_size個data
total_batch = (int)(mnist.train.num_examples/batch_size)
#start training
for epoch in range(traning_epochs):
for i in range(total_batch):
batch_xs, batch_ys = mnist.train.next_batch(batch_size) #max(x)=1,min(x)=0
_, c = sess.run([optimizer, cost], feed_dict={X:batch_xs})
if epoch % display_step == 0:
print("Epoch:", '%04d'%(epoch+1), "cost:", "{:.9f}".format(c))
print("Optimizer finished!")
#解壓前的結果
encoder_result = sess.run(encoder_op, feed_dict={X:mnist.test.images})
plt.scatter(encoder_result[:, 0], encoder_result[:, 1], c=mnist.test.labels)
plt.show()