Out-of-Sample Testing for GANs
This addresses the need for standardized evaluation in generative modeling, but it is incremental as it builds on existing GAN testing approaches.
The paper tackles the problem of evaluating GANs by proposing EvalGAN, a method that measures reconstruction quality and likelihood on test sets without auxiliary networks, and tests it on three state-of-the-art GANs with CIFAR-10 and CelebA datasets.
We propose a new method to evaluate GANs, namely EvalGAN. EvalGAN relies on a test set to directly measure the reconstruction quality in the original sample space (no auxiliary networks are necessary), and it also computes the (log)likelihood for the reconstructed samples in the test set. Further, EvalGAN is agnostic to the GAN algorithm and the dataset. We decided to test it on three state-of-the-art GANs over the well-known CIFAR-10 and CelebA datasets.