f-GAN-pytorch

A pytorch implementation of "f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization"

Result

• 1 epochs

  Kullback-Leibler	Reverse-KL	Pearson-X2
  Squared-Hellinger	Jensen-Shannon	GAN

• 20 epochs

  Kullback-Leibler	Reverse-KL	Pearson-X2
  Squared-Hellinger	Jensen-Shannon	GAN

  • 7 epochs Kullback-Leibler divergence
  

  Mode collapse occurs after 7 epochs for Kullback-Leibler divergence objective.
  This is the result of 7 epochs result of Kullback-Leibler divergence objective.