import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np

tf.set_random_seed(1)
np.random.seed(1)

BATCH_SIZE = 64
LR_G = 0.0001
LR_D = 0.0001
N_IDEAS = 5
ART_COMPONENTS = 15
PAINT_POINTS = np.vstack([np.linspace(-1,1,ART_COMPONENTS)for _ in range(BATCH_SIZE)])                                                #shape = (64,15)

print(PAINT_POINTS)

plt.plot(PAINT_POINTS[0],2*np.power(PAINT_POINTS[0],2)+1,c = '#74BCFF',lw = 3,label='upper bound')
plt.plot(PAINT_POINTS[0],1*np.power(PAINT_POINTS[0],2)+0,c = '#FF9359',lw = 3,label='lower bound')
plt.legend(loc = 'upper right')
plt.show()

def artist_works():                                            #即真實的資料
    a = np.random.uniform(1,2,size=BATCH_SIZE)[:,np.newaxis]   #shape = (64,1)
    paintings = a*np.power(PAINT_POINTS,2)+(a-1)               #shape = (64,15)
    return paintings

with tf.variable_scope('Generator'):                           #使用生成器偽造假的資料
    G_in = tf.placeholder(tf.float32,[None,N_IDEAS])           #shape = (64,5)
    G_l1 = tf.layers.dense(G_in,128,tf.nn.relu)
    G_out = tf.layers.dense(G_l1,ART_COMPONENTS)


with tf.variable_scope('Discriminator'):
    real_art = tf.placeholder(tf.float32,[None,ART_COMPONENTS],name='real_in')  #使用鑑別器來鑑別真實資料
    D_l0 = tf.layers.dense(real_art,128,tf.nn.relu,name='1')                    #並將它判別為1
    prob_artist0 = tf.layers.dense(D_l0,1,tf.nn.sigmoid,name='out')

    #fake art
    D_l1 = tf.layers.dense(G_out,128,tf.nn.relu,name='1',reuse=True)            #使用費鑑別器來判別偽造資料
    prob_artist1 = tf.layers.dense(D_l1,1,tf.nn.sigmoid,name='out',reuse=True)  #並將其判別為0

D_loss = -tf.reduce_mean(tf.log(prob_artist0)+tf.log(1-prob_artist1))           #定義誤差函式
G_loss = tf.reduce_mean(tf.log(1-prob_artist1))

train_D = tf.train.AdamOptimizer(LR_D).minimize(                                #定義優化函式
       D_loss,var_list=tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,scope='Discriminator'))

train_G = tf.train.AdamOptimizer(LR_G).minimize(
       G_loss,var_list=tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,scope='Generator'))

sess= tf.Session()                                                              #初始化流圖
sess.run(tf.global_variables_initializer())         

plt.ion()
for step in range(5000):
    artist_paintings = artist_works()
    G_ideas = np.random.randn(BATCH_SIZE,N_IDEAS)
    G_paintings,pa0,D1 = sess.run([G_out,prob_artist0,D_loss,train_D,train_G],
                                  {G_in:G_ideas,real_art:artist_paintings})[:3]

    if step%50==0:                                                               #視覺化
        plt.cla()
        plt.plot(PAINT_POINTS[0],G_paintings[0],c='#4AD631',lw=3,label='Generated painting')
        plt.plot(PAINT_POINTS[0],2*np.power(PAINT_POINTS[0],2)+1,c='#74BCFF',lw=3,label='upper bound')
        plt.plot(PAINT_POINTS[0],1*np.power(PAINT_POINTS[0],2)+0,c='#FF9359',lw=3,label='lower bound')
        plt.text(-.5,2.3,'D accuracy=%.2f (0.5 for D to converge)'%pa0.mean(),fontdict={'size':15})
        plt.text(-.5,2,'D score=%.2f (-1.38 for G to converge)'%-D1,fontdict={'size':15})
        plt.ylim((0,3));plt.legend(loc='upper right',fontsize=12);plt.draw();plt.pause(0.01)

plt.ioff()
plt.show()
 

本程式使用生成對抗網路（GAN）來實現一個小Demo：

目的 ：使用生成器擬合形如：a*X2+（a-1）

對於a 我們使用np.random.uniform函式隨機生成（1-2之間）

為了直觀的看到生成器生成的資料，我們設定了上下兩個邊界。

最終實驗效果：

註釋：此文為莫凡Python學習筆記 莫凡tensorflow