Abstract:Image generation is a hot topic in the field of machine learning and computer vision. As a representative of its algorithm, the Generative Adversarial Network (GAN) has the problem of mode collapse in practice. The proposed Dual Discriminator Weighted Mixture Generative Adversarial Network (D2WMGAN) approach can cope with this problem. On the one hand, the D2WMGAN uses the mixed distribution of multiple generators to approximate the real distribution, in order to prevent the extreme situation that multiple generators learn the same distribution and generate the same class of samples, with a classifier to play games with generators to make different generators learn different distributions. On the other hand, the objective function of D2WMGAN weights the Kullback–Leibler (KL) divergence and the reverse KL divergence, and uses their complementary characteristics to improve the quality and diversity of samples from the generators. Then, the theoretical conditional optimality of the D2WMGAN is proved theoretically, which shows that multiple generators can learn the real data distribution in the case of the optimal discriminator and classifier. Finally, extensive experiments are conducted on a large amount of synthetic data and real-world large-scale datasets (such as, CIFAR-10 and MNIST), and the commonly used GAN evaluation indicators (Wasserstein distance, JS divergence, Inception score, and Frechet Inception Distance) are introduced for comparative analysis. Experimental results show that the proposed D2WMGAN approach can better learn multiple mode data, generate rich realistic samples, and effectively solve the problem of mode collapse.

Normalized Diversification

Statistics Enhancement Generative Adversarial Networks for Diverse Conditional Image Synthesis

Diversity Regularized Adversarial Learning

Selectively increasing the diversity of GAN-generated samples

Towards the Gradient Vanishing, Divergence Mismatching and Mode Collapse of Generative Adversarial Nets

Alleviating Mode Collapse in GAN via Diversity Penalty Module

Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis

Mode Regularized Generative Adversarial Networks

CHAIN: Enhancing Generalization in Data-Efficient GANs via lipsCHitz continuity constrAIned Normalization

UniGAN: Reducing Mode Collapse in GANs using a Uniform Generator

Learning Semantic-aware Normalization for Generative Adversarial Networks.

Dual Discriminator Weighted Mixture Generative Adversarial Network for image generation

Adaptive Spectral Normalization for Generative Models

DivGAN: A Diversity Enforcing Generative Adversarial Network for Mode Collapse Reduction

Improving Generative Adversarial Networks via Adversarial Learning in Latent Space

Optimizing Generative Adversarial Networks in Latent Space

Generative Adversarial Networks Based on Denoising and Reconstruction Regularization

Spectral Normalization for Generative Adversarial Networks

Diversity-Promoting GAN: A Cross-Entropy Based Generative Adversarial Network for Diversified Text Generation

Optimizing Latent Distributions for Non-Adversarial Generative Networks

Utilizing Amari-Alpha Divergence to Stabilize the Training of Generative Adversarial Networks