MMCGAN: Generative Adversarial Network with Explicit Manifold Prior

TL;DR

MMC GAN introduces an explicit manifold prior based on Minimum Manifold Coding to address mode collapse and training instability in GANs, leading to more realistic and diverse generated samples.

Contribution

The paper proposes MMC as a novel manifold learning approach to improve GAN training stability and sample diversity by encouraging simple, unfolded manifolds.

Findings

MMC GAN reduces mode collapse in experiments.

Training stability is significantly improved.

Sample quality is enhanced with the proposed method.

Abstract

Generative Adversarial Network(GAN) provides a good generative framework to produce realistic samples, but suffers from two recognized issues as mode collapse and unstable training. In this work, we propose to employ explicit manifold learning as prior to alleviate mode collapse and stabilize training of GAN. Since the basic assumption of conventional manifold learning fails in case of sparse and uneven data distribution, we introduce a new target, Minimum Manifold Coding (MMC), for manifold learning to encourage simple and unfolded manifold. In essence, MMC is the general case of the shortest Hamiltonian Path problem and pursues manifold with minimum Riemann volume. Using the standardized code from MMC as prior, GAN is guaranteed to recover a simple and unfolded manifold covering all the training data. Our experiments on both the toy data and real datasets show the effectiveness of MMCGAN in alleviating mode collapse, stabilizing training, and improving the quality of generated samples.