WGAN 简介与代码实战
1.介绍
原始gan存在着训练困难、生成器和判别器的loss无法指示训练进程、生成样本缺乏多样性等问题。中间虽然有一序列gan想解决掉这些问题，但都是治标不治本的方案，直到WGAN的出现，并且作者从理论上证明怎么来解决这些问题，可见作者的数学功底是真的很强悍，更加详细的内容可参见论文：Wasserstein GAN
2.模型结构
整个算法流程，我们注意这两点就行：1. 优化器选择RMSProp； 2. 判别器的在更新参数的时候会被剪切到一个范围【-c， c】
3.模型特点
WGAN相比原始GAN的算法实现流程却只改了四点：
- 判别器最后一层去掉sigmoid （原始GAN的判别器做的是真假二分类任务，所以最后一层是sigmoid，但是现在WGAN中的判别器做的是近似拟合Wasserstein距离，属于回归任务，所以要把最后一层的sigmoid拿掉）
- 生成器和判别器的loss不取log （Ian Goodfellow提出的“- log D trick”）
- 每次更新判别器的参数之后把它们的绝对值截断到不超过一个固定常数c (用来满足lipschitz连续性条件)
- 不要用基于动量的优化算法（包括momentum和Adam），推荐RMSProp，SGD也行 （作者发现这样训练比较稳定）
4.代码实现 keras
class WGAN():
def __init__(self):
self.img_rows = 28
self.img_cols = 28
self.channels = 1
self.img_shape = (self.img_rows, self.img_cols, self.channels)
self.latent_dim = 100
# Following parameter and optimizer set as recommended in paper
self.n_critic = 5
self.clip_value = 0.01
optimizer = RMSprop(lr=0.00005)
# Build and compile the critic
self.critic = self.build_critic()
self.critic.compile(loss=self.wasserstein_loss,
optimizer=optimizer,
metrics=['accuracy'])
# Build the generator
self.generator = self.build_generator()
# The generator takes noise as input and generated imgs
z = Input(shape=(self.latent_dim,))
img = self.generator(z)
# For the combined model we will only train the generator
self.critic.trainable = False
# The critic takes generated images as input and determines validity
valid = self.critic(img)
# The combined model (stacked generator and critic)
self.combined = Model(z, valid)
self.combined.compile(loss=self.wasserstein_loss,
optimizer=optimizer,
metrics=['accuracy'])
def wasserstein_loss(self, y_true, y_pred):
return K.mean(y_true * y_pred)
def build_generator(self):
model = Sequential()
model.add(Dense(128 * 7 * 7, activation="relu", input_dim=self.latent_dim))
model.add(Reshape((7, 7, 128)))
model.add(UpSampling2D())
model.add(Conv2D(128, kernel_size=4, padding="same"))
model.add(BatchNormalization(momentum=0.8))
model.add(Activation("relu"))
model.add(UpSampling2D())
model.add(Conv2D(64, kernel_size=4, padding="same"))
model.add(BatchNormalization(momentum=0.8))
model.add(Activation("relu"))
model.add(Conv2D(self.channels, kernel_size=4, padding="same"))
model.add(Activation("tanh"))
model.summary()
noise = Input(shape=(self.latent_dim,))
img = model(noise)
return Model(noise, img)
def build_critic(self):
model = Sequential()
model.add(Conv2D(16, kernel_size=3, strides=2, input_shape=self.img_shape, padding="same"))
model.add(LeakyReLU(alpha=0.2))
model.add(Dropout(0.25))
model.add(Conv2D(32, kernel_size=3, strides=2, padding="same"))
model.add(ZeroPadding2D(padding=((0,1),(0,1))))
model.add(BatchNormalization(momentum=0.8))
model.add(LeakyReLU(alpha=0.2))
model.add(Dropout(0.25))
model.add(Conv2D(64, kernel_size=3, strides=2, padding="same"))
model.add(BatchNormalization(momentum=0.8))
model.add(LeakyReLU(alpha=0.2))
model.add(Dropout(0.25))
model.add(Conv2D(128, kernel_size=3, strides=1, padding="same"))
model.add(BatchNormalization(momentum=0.8))
model.add(LeakyReLU(alpha=0.2))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(1))
model.summary()
img = Input(shape=self.img_shape)
validity = model(img)
return Model(img, validity)
def train(self, epochs, batch_size=128, sample_interval=50):
# Load the dataset
(X_train, _), (_, _) = mnist.load_data()
# Rescale -1 to 1
X_train = (X_train.astype(np.float32) - 127.5) / 127.5
X_train = np.expand_dims(X_train, axis=3)
# Adversarial ground truths
valid = -np.ones((batch_size, 1))
fake = np.ones((batch_size, 1))
for epoch in range(epochs):
for _ in range(self.n_critic):
# ---------------------
# Train Discriminator
# ---------------------
# Select a random batch of images
idx = np.random.randint(0, X_train.shape[0], batch_size)
imgs = X_train[idx]
# Sample noise as generator input
noise = np.random.normal(0, 1, (batch_size, self.latent_dim))
# Generate a batch of new images
gen_imgs = self.generator.predict(noise)
# Train the critic
d_loss_real = self.critic.train_on_batch(imgs, valid)
d_loss_fake = self.critic.train_on_batch(gen_imgs, fake)
d_loss = 0.5 * np.add(d_loss_fake, d_loss_real)
# Clip critic weights
for l in self.critic.layers:
weights = l.get_weights()
weights = [np.clip(w, -self.clip_value, self.clip_value) for w in weights]
l.set_weights(weights)
# ---------------------
# Train Generator
# ---------------------
g_loss = self.combined.train_on_batch(noise, valid)
# Plot the progress
print ("%d [D loss: %f] [G loss: %f]" % (epoch, 1 - d_loss[0], 1 - g_loss[0]))
# If at save interval => save generated image samples
if epoch % sample_interval == 0:
self.sample_images(epoch)
def sample_images(self, epoch):
r, c = 5, 5
noise = np.random.normal(0, 1, (r * c, self.latent_dim))
gen_imgs = self.generator.predict(noise)
# Rescale images 0 - 1
gen_imgs = 0.5 * gen_imgs + 0.5
fig, axs = plt.subplots(r, c)
cnt = 0
for i in range(r):
for j in range(c):
axs[i,j].imshow(gen_imgs[cnt, :,:,0], cmap='gray')
axs[i,j].axis('off')
cnt += 1
fig.savefig("images/mnist_%d.png" % epoch)
plt.close()