Computes one of E-GAN's three named mutation objectives against a discriminator's own raw logit output, and its gradient w.r.t. those logits. Every objective trains the generator to make the discriminator's output move toward the "real" (1) class – they differ only in how that pressure is shaped (saturating vs. non-saturating vs. quadratic). More...

#include <egan_mutation.hpp>

Public Member Functions
	MutationLoss (DeviceBackend *backend)

float	Forward (MutationObjective objective, const Tensor &logits)
	Computes the chosen objective's scalar value against logits.

Tensor	Backward () const
	Gradient w.r.t. the logits passed to the most recent Forward() call.

Detailed Description

Computes one of E-GAN's three named mutation objectives against a discriminator's own raw logit output, and its gradient w.r.t. those logits. Every objective trains the generator to make the discriminator's output move toward the "real" (1) class – they differ only in how that pressure is shaped (saturating vs. non-saturating vs. quadratic).

Constructor & Destructor Documentation

◆ MutationLoss()

pulsatrix::MutationLoss::MutationLoss ( DeviceBackend * backend )

inlineexplicit

Member Function Documentation

◆ Backward()

Tensor pulsatrix::MutationLoss::Backward ( ) const

inline

Gradient w.r.t. the logits passed to the most recent Forward() call.

Exceptions

std::logic_error if Forward() has never been called (delegated from the underlying loss's own backward()).

◆ Forward()

float pulsatrix::MutationLoss::Forward	(	MutationObjective	objective,
		const Tensor &	logits
	)