Semi-Supervised Learning with GANs: Revisiting Manifold Regularization

Abstract

GANS are powerful generative models that are able to model the manifold of natural images. We leverage this property to perform manifold regularization by approximating the Laplacian norm using a Monte Carlo approximation that is easily computed with the GAN. When incorporated into the feature-matching GAN of Improved GAN, we achieve state-of-the-art results for GAN-based semi-supervised learning on the CIFAR-10 dataset, with a method that is significantly easier to implement than competing methods.

arXiv commentAccepted paper

Cited by in corpus (6)

• Semi-supervised learning based on generative adversarial network: a comparison between good GAN and bad GAN approach
• Adversarial Partial Multi-Label Learning
• MR-GAN: Manifold Regularized Generative Adversarial Networks
• Semi-supervised Learning using Adversarial Training with Good and Bad Samples
• Patch-level Neighborhood Interpolation: A General and Effective Graph-based Regularization Strategy
• Adversarial Code Learning for Image Generation