Chainer implementation of the Wasserstein GAN by Martin Arjovsky et al. Note that this is not the official implementation. The official implementation is https://github.com/martinarjovsky/WassersteinGAN.
Also, a summary of the paper can be found here. It explains the intuition behind the approximation of the EM distance and the problem with the Jensen-Shannon divergence.
Train the models with CIFAR-10. Images will be randomly sampled from the generator after each epoch, and saved under a subdirectory result/ (which is created automatically).
python train.py --batch-size 64 --epochs 100 --gpu 1Plotting the estimates with CIFAR-10.
