Dual Discriminator Generative Adversarial Nets

Abstract

We propose in this paper a novel approach to tackle the problem of mode collapse encountered in generative adversarial network (GAN). Our idea is intuitive but proven to be very effective, especially in addressing some key limitations of GAN. In essence, it combines the Kullback-Leibler (KL) and reverse KL divergences into a unified objective function, thus it exploits the complementary statistical properties from these divergences to effectively diversify the estimated density in capturing multi-modes. We term our method dual discriminator generative adversarial nets (D2GAN) which, unlike GAN, has two discriminators; and together with a generator, it also has the analogy of a minimax game, wherein a discriminator rewards high scores for samples from data distribution whilst another discriminator, conversely, favoring data from the generator, and the generator produces data to fool both two discriminators. We develop theoretical analysis to show that, given the maximal discriminators, optimizing the generator of D2GAN reduces to minimizing both KL and reverse KL divergences between data distribution and the distribution induced from the data generated by the generator, hence effectively avoiding the mode collapsing problem. We conduct extensive experiments on synthetic and real-world large-scale datasets (MNIST, CIFAR-10, STL-10, ImageNet), where we have made our best effort to compare our D2GAN with the latest state-of-the-art GAN's variants in comprehensive qualitative and quantitative evaluations. The experimental results demonstrate the competitive and superior performance of our approach in generating good quality and diverse samples over baselines, and the capability of our method to scale up to ImageNet database.

References in corpus (13)

• Adam: A Method for Stochastic Optimization
• Conditional Generative Adversarial Nets
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks
• Improved Techniques for Training GANs
• BEGAN: Boundary Equilibrium Generative Adversarial Networks
• Adversarially Learned Inference
• A note on the evaluation of generative models
• Mode Regularized Generative Adversarial Networks
• How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?
• MAGAN: Margin Adaptation for Generative Adversarial Networks
• Multi-Generator Generative Adversarial Nets
• Geometric Enclosing Networks

Cited by in corpus (21)

• TransGAN: Two Pure Transformers Can Make One Strong GAN, and That Can Scale Up
• A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications
• Cross-view image synthesis using geometry-guided conditional GANs
• Multi-Generator Generative Adversarial Nets
• Photographic Text-to-Image Synthesis with a Hierarchically-nested Adversarial Network
• Dropout-GAN: Learning from a Dynamic Ensemble of Discriminators
• MGGAN: Solving Mode Collapse using Manifold Guided Training
• Audio inpainting with generative adversarial network
• Stackelberg GAN: Towards Provable Minimax Equilibrium via Multi-Generator Architectures
• cGANs with Multi-Hinge Loss
• Comparing Generative Adversarial Network Techniques for Image Creation and Modification
• Training Generative Adversarial Networks via Primal-Dual Subgradient Methods: A Lagrangian Perspective on GAN
• Self-Supervised GANs with Label Augmentation
• Understanding and Stabilizing GANs' Training Dynamics with Control Theory
• GestureGAN for Hand Gesture-to-Gesture Translation in the Wild
• Adaptive Weighted Discriminator for Training Generative Adversarial Networks
• Generation For Adaption: A GAN-Based Approach for 3D Domain Adaption with Point Cloud Data
• Structured Latent Embeddings for Recognizing Unseen Classes in Unseen Domains
• Student's t-Generative Adversarial Networks
• Restrained Generative Adversarial Network against Overfitting in Numeric Data Augmentation
• KGAN: How to Break The Minimax Game in GAN