Multi-Generator Generative Adversarial Nets

Hoang, Quan; Nguyen, Tu Dinh; Le, Trung; Phung, Dinh

doi:10.48550/arxiv.1708.02556

Cited by 24 publications

(32 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It contains 10 balanced classes of images of different objects. Just to further illustrate the merits of multi-agent GANs, we offer a selection of images generated by WGAN-GP [ACB17] and the MGAN [HNLP17].…”

Section: Cifar-10mentioning

confidence: 99%

Teamwork makes von Neumann work: Min-Max Optimization in Two-Team Zero-Sum Games

Fivos¹,

Panageas²,

Vlatakis-Gkaragkounis³

2021

Preprint

View full text Add to dashboard Cite

Motivated by recent advances in both theoretical and applied aspects of multiplayer games, spanning from e-sports to multi-agent generative adversarial networks, we focus on min-max optimization in team zero-sum games. In this class of games, players are split into two teams with payoffs equal within the same team and of opposite sign across the opponent team. Unlike the textbook two-player zero-sum games, finding a Nash equilibrium in our class can be shown to be CLS-hard, i.e., it is unlikely to have a polynomial-time algorithm for computing Nash equilibria. Moreover, in this generalized framework, we establish that even asymptotic last iterate or time average convergence to a Nash Equilibrium is not possible using Gradient Descent Ascent (GDA), its optimistic variant, and extra gradient. Specifically, we present a family of team games whose induced utility is non multi-linear with non attractive per-se mixed Nash Equilibria, as strict saddle points of the underlying optimization landscape. Leveraging techniques from control theory, we complement these negative results by designing a modified GDA that converges locally to Nash equilibria. Finally, we discuss connections of our framework with AI architectures with team competition structures like multi-agent generative adversarial networks.

show abstract

Section: Cifar-10mentioning

confidence: 99%

Teamwork makes von Neumann work: Min-Max Optimization in Two-Team Zero-Sum Games

Fivos¹,

Panageas²,

Vlatakis-Gkaragkounis³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…It is worth mentioning that generators of the above three GAN models are not trained simultaneously, which may cover more modes as well as more poor samples step by step. Ensemble GAN models that train generators simultaneously include Mix+GAN (Arora et al, 2017), MAD-GAN (Ghosh et al, 2018), MGAN (Hoang et al, 2017), MEGAN (Park et al, 2018), andDeLiGAN (Gurumurthy et al, 2017). Mix+GAN has multiple generators and discriminators that have independent neural network parameters and learnable mixed weights, which is computational expensive.…”

Section: Related Workmentioning

confidence: 99%

“…MGAN also adds extra classification loss term in the generator's loss function to force each generator to specialize on different modes in the training set. (Hoang et al, 2017) pointed out that the mixture weights of the generator distributions of MGAN are unreasonably fixed and evenly distributed, and proposed Mixture of Experts GAN (MEGAN). Inside the MEGAN, a learnable Gating Networks based on Straight-Through Gumbel Softmax (Jang et al, 2017) picks one sample from all the generated samples of multiple generators as the output of the model.…”

Section: Related Workmentioning

confidence: 99%

“…To improve the generator distribution, many methods have been proposed and they can be roughly divided into three categories: (1) Improving the training approach, such as Unrolled GAN (Metz et al, 2017) or WGAN-GP (Gulrajani et al, 2017), etc., to overcome the problem of mode dropping by stabilizing GAN's training; (2) Using label conditioning, such as the conditional LAPGAN (Denton et al, 2015), AC-GAN (Odena et al, 2017), and cGAN (Mirza & Osindero, 2014), which can always significantly improve the sample quality. (Goodfellow, 2016;Salimans et al, 2016); (3) Ensembling multiple GANs, such as MGAN (Hoang et al, 2017), AdaGAN (Tolstikhin et al, 2017), Mix+GAN (Arora et al, 2017), and MAD-GAN (Ghosh et al, 2018), to cover more modes and improve the fidelity of generator distribution. However, conditional GANs can not be trained on unlabeled data sets.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Virtual Conditional Generative Adversarial Networks

Shi¹,

Cai²,

Shang³

et al. 2019

Preprint

View full text Add to dashboard Cite

When trained on multimodal image datasets, normal Generative Adversarial Networks (GANs) are usually outperformed by class-conditional GANs and ensemble GANs, but conditional GANs is restricted to labeled datasets and ensemble GANs lack efficiency. We propose a novel GAN variant called virtual conditional GAN (vcGAN) which is not only an ensemble GAN with multiple generative paths while adding almost zero network parameters, but also a conditional GAN that can be trained on unlabeled datasets without explicit clustering steps or objectives other than the adversary loss. Inside the vcGANs generator, a learnable "analog-to-digital converter (ADC)" module maps a slice of the inputted multivariate Gaussian noise to discrete/digital noise (virtual label), according to which a selector selects the corresponding generative path to produce the sample. All the generative paths share the same decoder network while in each path the decoder network is fed with a concatenation of a different pre-computed amplified one-hot vector and the inputted Gaussian noise. We conducted a lot of experiments on several balanced/imbalanced image datasets to demonstrate that vcGAN converges faster and achieves improved Frecht Inception Distance (FID). In addition, we show the training byproduct that the ADC in vcGAN learned the categorical probability of each mode and that each generative path generates samples of specific mode, which enables class-conditional sampling. Codes are available at https:// github.com/annonnymmouss/vcgan

show abstract

“…Albuquerque et al [1] show that training GAN variants with multiple discriminators is a practical approach even though extra capacity and computational cost are needed. Employing multiple generators and one discriminator to overcome the mode collapse issue and encourages diverse images has also been proposed [17,11].…”

Section: Introductionmentioning

confidence: 99%

DuelGAN: A Duel Between Two Discriminators Stabilizes the GAN Training

Wei¹,

Liu²,

Luo³

et al. 2021

Preprint

View full text Add to dashboard Cite

In this paper, we introduce PeerGAN, a generative adversarial network (GAN) solution to improve the stability of the generated samples and to mitigate mode collapse. Built upon the Vanilla GAN's two-player game between the discriminator D 1 and the generator G, we introduce a peer discriminator D 2 to the min-max game. Similar to previous work using two discriminators, the first role of both D 1 , D 2 is to distinguish between generated samples and real ones, while the generator tries to generate high-quality samples which are able to fool both discriminators. Different from existing methods, we introduce another game between D 1 and D 2 to discourage their agreement and therefore increase the level of diversity of the generated samples. This property helps avoid early mode collapse by preventing D 1 and D 2 from converging too fast. We provide theoretical analysis for the equilibrium of the min-max game formed among G, D 1 , D 2 . We offer convergence behavior of PeerGAN as well as stability of the min-max game. It's worth mentioning that PeerGAN operates in the unsupervised setting, and the additional game between D 1 and D 2 does not need any label supervision. Experiments results on a synthetic dataset and on real-world image datasets (MNIST, Fashion MNIST, CIFAR-10, STL-10, CelebA, VGG) demonstrate that PeerGAN outperforms competitive baseline work in generating diverse and high-quality samples, while only introduces negligible computation cost.

show abstract

Multi-Generator Generative Adversarial Nets

Cited by 24 publications

References 2 publications

Teamwork makes von Neumann work: Min-Max Optimization in Two-Team Zero-Sum Games

Teamwork makes von Neumann work: Min-Max Optimization in Two-Team Zero-Sum Games

Virtual Conditional Generative Adversarial Networks

DuelGAN: A Duel Between Two Discriminators Stabilizes the GAN Training

Contact Info

Product

Resources

About