Generalized Loss-Sensitive Adversarial Learning With Manifold Margins

Keywords

Image generation; Lipschitz regularization; Regularized GAN; Semi-supervised classification

Abstract

The classic Generative Adversarial Net and its variants can be roughly categorized into two large families: the unregularized versus regularized GANs. By relaxing the non-parametric assumption on the discriminator in the classic GAN, the regularized GANs have better generalization ability to produce new samples drawn from the real distribution. It is well known that the real data like natural images are not uniformly distributed over the whole data space. Instead, they are often restricted to a low-dimensional manifold of the ambient space. Such a manifold assumption suggests the distance over the manifold should be a better measure to characterize the distinct between real and fake samples. Thus, we define a pullback operator to map samples back to their data manifold, and a manifold margin is defined as the distance between the pullback representations to distinguish between real and fake samples and learn the optimal generators. We justify the effectiveness of the proposed model both theoretically and empirically.

Publication Date

1-1-2018

Publication Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Volume

11209 LNCS

Number of Pages

90-104

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

DOI Link

https://doi.org/10.1007/978-3-030-01228-1_6

Socpus ID

85055122724 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85055122724

This document is currently not available here.

Share

COinS