Manifold regularization with GANs for semi-supervised learning
This addresses semi-supervised learning for image classification, offering an incremental improvement in GAN-based methods.
The paper tackles semi-supervised learning by using GANs for manifold regularization, achieving state-of-the-art results on CIFAR-10 and SVHN benchmarks with a simpler implementation than competitors.
Generative Adversarial Networks are powerful generative models that are able to model the manifold of natural images. We leverage this property to perform manifold regularization by approximating a variant of the Laplacian norm using a Monte Carlo approximation that is easily computed with the GAN. When incorporated into the semi-supervised feature-matching GAN we achieve state-of-the-art results for GAN-based semi-supervised learning on CIFAR-10 and SVHN benchmarks, with a method that is significantly easier to implement than competing methods. We also find that manifold regularization improves the quality of generated images, and is affected by the quality of the GAN used to approximate the regularizer.