accelbrainbase.computableloss._mxnet.discriminatorloss package

Submodules

accelbrainbase.computableloss._mxnet.discriminatorloss.eb_discriminator_loss module

class accelbrainbase.computableloss._mxnet.discriminatorloss.eb_discriminator_loss.EBDiscriminatorLoss(weight=None, batch_axis=0, margin=1.0, margin_decay_rate=0.1, margin_decay_epoch=50, **kwargs)

Bases: accelbrainbase.computableloss._mxnet.discriminator_loss.DiscriminatorLoss

Loss function of discriminators in Generative Adversarial Networks(GANs).

compute(true_posterior_arr, generated_posterior_arr)

Compute loss.

Parameters:
  • true_posterior_arr – Real samples.
  • generated_posterior_arr – Generated samples.
Returns:

Tensor of losses.
hybrid_forward(F, true_posterior_arr, generated_posterior_arr, sample_weight=None)

Forward propagation, computing losses.

Parameters:
  • F – mxnet.ndarray or mxnet.symbol.
  • true_posterior_arr – mxnet.ndarray or mxnet.symbol of true posterior inferenced by the discriminator.
  • generated_posterior_arr – mxnet.ndarray or mxnet.symbol of fake posterior inferenced by the generator.
Returns:

mxnet.ndarray or mxnet.symbol of loss.

Module contents