pygan.gansvaluefunction package

Submodules

pygan.gansvaluefunction.margin_loss module

class pygan.gansvaluefunction.margin_loss.MarginLoss(margin=1.0, margin_attenuate_rate=0.1, attenuate_epoch=50)

Bases: pygan.gans_value_function.GANsValueFunction

Value function in energy-based GANs framework.

References

• Zhao, J., Mathieu, M., & LeCun, Y. (2016). Energy-based generative adversarial network. arXiv preprint arXiv:1609.03126.

compute_discriminator_reward(true_posterior_arr, generated_posterior_arr)

Compute discriminator’s reward.

Parameters:

• true_posterior_arr – np.ndarray of true posterior inferenced by the discriminator.
• generated_posterior_arr – np.ndarray of fake posterior inferenced by the discriminator.

Returns:

np.ndarray of Gradients.

compute_generator_reward(generated_posterior_arr)

Compute generator’s reward.

Parameters:generated_posterior_arr – np.ndarray of fake posterior inferenced by the discriminator. Returns:np.ndarray of Gradients.

discriminator_reward_arr

getter

get_discriminator_reward_arr()

getter

get_margin()

getter

margin

getter

set_readonly(value)

setter

pygan.gansvaluefunction.mini_max module

class pygan.gansvaluefunction.mini_max.MiniMax

Bases: pygan.gans_value_function.GANsValueFunction

Value function in GANs framework.

compute_discriminator_reward(true_posterior_arr, generated_posterior_arr)

Compute discriminator’s reward.

Parameters:

• true_posterior_arr – np.ndarray of true posterior inferenced by the discriminator.
• generated_posterior_arr – np.ndarray of fake posterior inferenced by the discriminator.

Returns:

np.ndarray of Gradients.

compute_generator_reward(generated_posterior_arr)

Compute generator’s reward.

Parameters:generated_posterior_arr – np.ndarray of fake posterior inferenced by the discriminator. Returns:np.ndarray of Gradients.

Module contents