glitchgan.tf.gan_models
=======================

.. py:module:: glitchgan.tf.gan_models

.. autoapi-nested-parse::

   TensorFlow/Keras GAN model classes for cDVGAN.

   Includes:
   - cWGAN   — conditional Wasserstein GAN with gradient penalty
   - cDVGAN  — adds a first-derivative discriminator
   - cDVGAN2 — adds first and second derivative discriminators
   - build_gan() — factory function









Module Contents
---------------

.. py:data:: NUM_CLASSES
   :value: 7


.. py:class:: cWGAN(signal_length=8192, num_classes=NUM_CLASSES, noise_dim=100, d_steps=5, gp_weight=10.0, lr=0.0001)

   Bases: :py:obj:`keras.Model`


   Conditional Wasserstein GAN with gradient penalty.


   .. py:attribute:: noise_dim
      :value: 100



   .. py:attribute:: num_classes
      :value: 7



   .. py:attribute:: d_steps
      :value: 5



   .. py:attribute:: gp_weight
      :value: 10.0



   .. py:attribute:: generator


   .. py:attribute:: discriminator


   .. py:attribute:: g_optimizer


   .. py:attribute:: d_optimizer


   .. py:method:: train_step(data)


.. py:class:: cDVGAN(signal_length=8192, num_classes=NUM_CLASSES, noise_dim=100, d_steps=5, gp_weight=10.0, lr=0.0001)

   Bases: :py:obj:`keras.Model`


   Conditional Dual-discriminator Variational GAN (first derivative).


   .. py:attribute:: noise_dim
      :value: 100



   .. py:attribute:: num_classes
      :value: 7



   .. py:attribute:: d_steps
      :value: 5



   .. py:attribute:: gp_weight
      :value: 10.0



   .. py:attribute:: generator


   .. py:attribute:: discriminator


   .. py:attribute:: deriv_discriminator


   .. py:attribute:: g_optimizer


   .. py:attribute:: d_optimizer


   .. py:attribute:: d2d_optimizer


   .. py:method:: train_step(data)


.. py:class:: cDVGAN2(signal_length=8192, num_classes=NUM_CLASSES, noise_dim=100, d_steps=5, gp_weight=10.0, lr=0.0001)

   Bases: :py:obj:`keras.Model`


   cDVGAN with an additional second-derivative discriminator.


   .. py:attribute:: noise_dim
      :value: 100



   .. py:attribute:: num_classes
      :value: 7



   .. py:attribute:: d_steps
      :value: 5



   .. py:attribute:: gp_weight
      :value: 10.0



   .. py:attribute:: generator


   .. py:attribute:: discriminator


   .. py:attribute:: deriv_discriminator


   .. py:attribute:: deriv2_discriminator


   .. py:attribute:: g_optimizer


   .. py:attribute:: d_optimizer


   .. py:attribute:: d2d_optimizer


   .. py:attribute:: d2d2_optimizer


   .. py:method:: train_step(data)


.. py:class:: GlitchGAN(noise_dim=100, d_steps=5, gp_weight=10.0, lr=0.0001)

   Bases: :py:obj:`cDVGAN`


   cDVGAN trained on LIGO gravitational-wave glitch data.

   Fixes the LIGO-specific defaults (signal length, number of glitch classes)
   so they don't need to be passed at every call site.  All architecture and
   training logic lives in :class:`cDVGAN`.


   .. py:attribute:: SIGNAL_LENGTH
      :value: 8192



   .. py:attribute:: NUM_CLASSES
      :value: 7



.. py:function:: build_gan(variant='cDVGAN', signal_length=8192, num_classes=NUM_CLASSES, noise_dim=100, d_steps=5, gp_weight=10.0, lr=0.0001)

   Instantiate a TF GAN variant by name.

   :param variant: One of ``"cWGAN"``, ``"cDVGAN"``, ``"cDVGAN2"``.
   :type variant: str

   :rtype: keras.Model