Missed TensorFlow World? Check out the recap. Learn more

tf.contrib.gan.GANModel

Class GANModel

Defined in tensorflow/contrib/gan/python/namedtuples.py.

A GANModel contains all the pieces needed for GAN training.

Generative Adversarial Networks (https://arxiv.org/abs/1406.2661) attempt to create an implicit generative model of data by solving a two agent game. The generator generates candidate examples that are supposed to match the data distribution, and the discriminator aims to tell the real examples apart from the generated samples.

Args:

  • generator_inputs: The random noise source that acts as input to the generator.
  • generated_data: The generated output data of the GAN.
  • generator_variables: A list of all generator variables.
  • generator_scope: Variable scope all generator variables live in.
  • generator_fn: The generator function.
  • real_data: A tensor or real data.
  • discriminator_real_outputs: The discriminator's output on real data.
  • discriminator_gen_outputs: The discriminator's output on generated data.
  • discriminator_variables: A list of all discriminator variables.
  • discriminator_scope: Variable scope all discriminator variables live in.
  • discriminator_fn: The discriminator function.

__new__

__new__(
    _cls,
    generator_inputs,
    generated_data,
    generator_variables,
    generator_scope,
    generator_fn,
    real_data,
    discriminator_real_outputs,
    discriminator_gen_outputs,
    discriminator_variables,
    discriminator_scope,
    discriminator_fn
)

Create new instance of GANModel(generator_inputs, generated_data, generator_variables, generator_scope, generator_fn, real_data, discriminator_real_outputs, discriminator_gen_outputs, discriminator_variables, discriminator_scope, discriminator_fn)

Properties

generator_inputs

generated_data

generator_variables

generator_scope

generator_fn

real_data

discriminator_real_outputs

discriminator_gen_outputs

discriminator_variables

discriminator_scope

discriminator_fn