小编pyt*_*way的帖子

GAN训练的难点

我正在尝试训练 GAN 以了解事件中多个特征的分布。训练的鉴别器和生成器的损失较低，但生成的事件具有不同形状的分布，但我不确定为什么。

我将 GAN 定义如下：

def create_generator():

    generator = Sequential()

    generator.add(Dense(50,input_dim=noise_dim))
    generator.add(LeakyReLU(0.2))    
    generator.add(Dense(25))
    generator.add(LeakyReLU(0.2))
    generator.add(Dense(5))
    generator.add(LeakyReLU(0.2))
    generator.add(Dense(len(variables), activation='tanh'))

    return generator


def create_descriminator():
    discriminator = Sequential()

    discriminator.add(Dense(4, input_dim=len(variables)))
    discriminator.add(LeakyReLU(0.2))
    discriminator.add(Dense(4))
    discriminator.add(LeakyReLU(0.2))
    discriminator.add(Dense(4))
    discriminator.add(LeakyReLU(0.2))
    discriminator.add(Dense(1, activation='sigmoid'))   
    discriminator.compile(loss='binary_crossentropy', optimizer=optimizer)
    return discriminator


discriminator = create_descriminator()
generator = create_generator()

def define_gan(generator, discriminator):
    # make weights in the discriminator not trainable
    discriminator.trainable = False
    model = Sequential()
    model.add(generator)
    model.add(discriminator)
    model.compile(loss = 'binary_crossentropy', optimizer=optimizer)
    return model

gan = define_gan(generator, discriminator)

Run Code Online (Sandbox Code Playgroud)

我使用这个循环训练 GAN：