Generative Multi-Adversarial Networks

By Ishan Durugkar, Ian Gemp, Sridhar Mahadevan. 2016

368 citations recorded by the source corpus.

Generative adversarial networks (GANs) are a framework for producing a generative model by way of a two-player minimax game. In this paper, we propose the \emph{Generative Multi-Adversarial Network} (GMAN), a framework that extends GANs to multiple discriminators. In previous work, the successful training of GANs requires modifying the minimax objective to accelerate training early on. In contrast, GMAN can be reliably trained with the original, untampered objective. We explore a number of design perspectives with the discriminator role ranging from formidable adversary to forgiving teacher. Image generation tasks comparing the proposed framework to standard GANs demonstrate GMAN produces higher quality samples in a fraction of the iterations when measured by a pairwise GAM-type metric.

Subfields: Computer Vision and Pattern Recognition, Statistical and Nonlinear Physics, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Statistical and Nonlinear Physics

In Generative Models

Citations in the corpus

A Style-Based Generator Architecture for Generative Adversarial Networks (2018, methodology): “, using multiple discriminators [17, 45], multiresolution discrimination [58, 53], or self-attention [61].”
Projected GANs Converge Faster (2021, background): “A different line of work considers a setup with multiple discriminators, applied to either the generated RGB image [8, 13] or low-dimensional projections thereof [1, 40].”
A U-Net Based Discriminator for Generative Adversarial Networks (2020, background): “Plenty of efforts have been made to improve the training and performance of GANs, from reformulation of the objective function [31, 1, 26, 37], integration of different regularization techniques [51, 34, 40, 48] and architectural changes [38, 19, 13, 27].”
High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs (2017, background): “n Eq. (1) then becomes a multi-task learning problem of min G max D1;D2;D3 X k=1;2;3 LGAN(G;D k): (3) Using multiple GAN discriminators at the same image scale has been proposed in unconditional GANs [12]. Iizuka et al. [20] add a global image classiﬁer to conditional GANs to synthesize globally coh”
StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks (2017, methodology): “Inspired by [10], we also build a baseline model with multiple discriminators at the 256×256 scale, namely “StackGANv2-3G3”.”
MoCoGAN: Decomposing Motion and Content for Video Generation (2017, background): “We note that, in a recent work [9], a GAN framework for image generation with mul-1 tiple discriminative networks was proposed.”
Multi-Adversarial Domain Adaptation (2018, background): “In particular, Generative Multi-Adversarial Network (GMAN) (Durugkar, Gemp, and Mahadevan 2017) extends GANs to multiple discriminators including formidable adversary and forgiving teacher, which signiﬁcantly eases model training and enhances distribution matching.”
Multi-agent Diverse Generative Adversarial Networks (2017, background): “Durugkar et al. [7] proposed a model with multiple discriminators whereby an ensemble of multiple discriminators have been shown to stabilize the training of the generator by guiding it to produce better samples.”
Photographic Text-to-Image Synthesis with a Hierarchically-Nested Adversarial Network (2018, methodology): “Particularly, using deep supervision [23] [10, 29].”
Pros and Cons of GAN Evaluation Measures (2018, background): “A variation of GAM known as generative multi-adversarial metric (GMAM), that is amenable to training with multiple discriminators, has been proposed in [21].”
IntroVAE: Introspective Variational Autoencoders for Photographic Image Synthesis (2018, methodology): “Some studies [11, 39] have trained a single generator with multiple discriminators ∗Ran He is the corresponding author.”
Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis (2019, methodology): “A number of methods focus on discriminators by introducing different divergence metrics [1, 18] and optimization process [6, 19, 24].”
Baby steps towards few-shot learning with multiple semantics (2019, methodology): “…14]; (iv) rendering synthetic examples [9, 35, 48]; (v) augmenting the training examples using geometric and photometric transformations [23] or learning adaptive augmentation strategies [17]; (vi) example synthesis using Generative Adversarial Networks (GANs) [2, 10, 16, 20, 22, 31, 37, 39, 59].”
MGAN: Training Generative Adversarial Nets with Multiple Generators (2018, background): “(Arjovsky et al., 2017) 3.82 ± 0.06 – – MIX+WGAN (Arora et al., 2017) 4.04 ± 0.07 – – Improved-GAN (Salimans et al., 2016) 4.36 ± 0.04 – – ALI (Dumoulin et al., 2016) 5.34 ± 0.05 – – BEGAN (Berthelot et al., 2017) 5.62 – – MAGAN (Wang et al., 2017) 5.67 – – GMAN (Durugkar et al., 2016) 6.00 ± 0.19 –”
A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT (2023, methodology): “To address mode collapse and improve the generative model, both the D2GAN [80] and GMAN [81] methods extend the traditional GANs by combining extra discriminators.”
Evolutionary Generative Adversarial Networks (2018, background): “Multidiscriminator GAN frameworks [46], [47] are devised and utilized for providing stable gradients to the generator and further stabilizing the adversarial training process.”
Neural Response Generation via GAN with an Approximate Embedding Layer (2017, background): “On the other hand, if the capacity of the discriminator is too high (such as “Filter[1, 2, 3]”), in the adversarial training, the training of the discriminator will converge too fast before the generator being sufﬁciently trained (Durugkar et al., 2016).”