GAN 生成对抗网络,新手画家和新手鉴赏家,自己是高级鉴赏家,告诉新手鉴赏家,做的画是否是高级画家的作品,新手画家和鉴赏家一起学习,一起对抗,一起成长。
import torch
import torchvision # 包括了数据库和图片的数据库
from torch.autograd import Variable
import torch.nn as nn
import matplotlib.pyplot as plt
import torch.utils.data as Data
import numpy as np
# Hyper Parameters 超参数
BATCH_SIZE = 64
LR_G = 0.0001 # learning rate for generator 新手画家
LR_D = 0.0001 # learning rate for discrimaniter 新手鉴赏家,需要学习才能知道画作是否好看
N_IDEAS = 5 # 灵感的个数
ART_COMPONENTS = 15
PAINT_POINTS = np.vstack([np.linspace(-1,1,ART_COMPONENTS) for _ in range(BATCH_SIZE)])
def artist_works():
a = np.random.uniform(1,2,size=BATCH_SIZE)[:,np.newaxis]
paintings = a * np.power(PAINT_POINTS,2) + (a-1) # 二次曲线,画笔即在蓝色线和红色线之间
paintings = torch.from_numpy(paintings).float()
return Variable(paintings)
# show our beautiful painting range
# 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()
G = nn.Sequential(
nn.Linear(N_IDEAS,128), # 用灵感创造一幅画
nn.ReLU(),
nn.Linear(128,ART_COMPONENTS),
)
D = nn.Sequential(
nn.Linear(ART_COMPONENTS,128),
nn.ReLU(),
nn.Linear(128,1),
nn.Sigmoid(), # 转换为一个百分比的形式
)
opt_G = torch.optim.Adam(G.parameters(),lr = LR_G)
opt_D = torch.optim.Adam(D.parameters(),lr = LR_D)
plt.ion()
for step in range(10000):# 学习1w 步
artist_paintings = artist_works() # 著名画家的画作
G_ideas = Variable(torch.randn(BATCH_SIZE,N_IDEAS))# 随机灵感,一个batch里面有N_IDEAS个数据
G_paintings = G(G_ideas) # 新手画家的画作
# 两幅画分别有多少是著名画家画的
prob_artist0 = D(artist_paintings) # 著名画家的概率
prob_artist1 = D(G_paintings) # 认为多少概率是从著名画家的画中学习的
# 尽可能增加认为是著名画家的概率,减少认为是新手画家的画
# 因为在tensorflow还有pytorch中需要最小化误差,因此加个负号
# 增加反而是负号
D_loss = -torch.mean(torch.log(prob_artist0) + torch.log(1. - prob_artist1))
# 减少反而是正号
G_loss = torch.mean(torch.log(1. - prob_artist1))
opt_D.zero_grad()
D_loss.backward(retain_graph=True)# 保留之前网络中的一些参数,下一次后向传递的时候再进行数据的传播
opt_D.step()
opt_G.zero_grad()
G_loss.backward()
opt_G.step()
if step % 50 == 0: # plotting
plt.cla()
plt.plot(PAINT_POINTS[0], G_paintings.data.numpy()[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)' % prob_artist0.data.numpy().mean(), fontdict={'size': 13})
plt.text(-.5, 2, 'D score= %.2f (-1.38 for G to converge)' % -D_loss.data.numpy(), fontdict={'size': 13})
plt.ylim((0, 3));plt.legend(loc='upper right', fontsize=10);plt.draw();plt.pause(0.01)
plt.ioff()
plt.show()