Implementation of semi-supervised GAN on MNIST dataset with 20% labeled data. The discriminator loss includes Supervised classification loss on labels and GAN loss of Real/Fake on Unsupervised data. Generator loss includes GAN loss along with feature matching loss as proposed in Improved Techniques for Training GANs: https://arxiv.org/pdf/1606.03498.pdf
namapolo / semisuper_gan Goto Github PK
View Code? Open in Web Editor NEWThis project forked from raghav64/semisuper_gan
Tf implementation of Semi-supervised GAN
Home Page: https://towardsdatascience.com/semi-supervised-learning-and-gans-f23bbf4ac683
License: MIT License