Triple Generative Adversarial Nets

By Chongxuan Li, Taufik Xu, Jun Zhu, Bo Zhang. 2017

475 citations recorded by the source corpus.

Generative Adversarial Nets (GANs) have shown promise in image generation and semi-supervised learning (SSL). However, existing GANs in SSL have two problems: (1) the generator and the discriminator (i.e. the classifier) may not be optimal at the same time; and (2) the generator cannot control the semantics of the generated samples. The problems essentially arise from the two-player formulation, where a single discriminator shares incompatible roles of identifying fake samples and predicting labels and it only estimates the data without considering the labels. To address the problems, we present triple generative adversarial net (Triple-GAN), which consists of three players---a generator, a discriminator and a classifier. The generator and the classifier characterize the conditional distributions between images and labels, and the discriminator solely focuses on identifying fake image-label pairs. We design compatible utilities to ensure that the distributions characterized by the classifier and the generator both converge to the data distribution. Our results on various datasets demonstrate that Triple-GAN as a unified model can simultaneously (1) achieve the state-of-the-art classification results among deep generative models, and (2) disentangle the classes and styles of the input and transfer smoothly in the data space via interpolation in the latent space class-conditionally.

Subfields: Computer Vision and Pattern Recognition, 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 Generative Models

Citations in the corpus

BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling (2019, neutral): “07 ) T RIPLE GAN [29] 0 .”
ORDisCo: Effective and Efficient Usage of Incremental Unlabeled Data for Semi-supervised Continual Learning (2021, methodology): “4, the predictions on the conditional samples of ORDisCo are significantly better than the one of a sequentially trained conditional GAN with SMB.”
Lifelong GAN: Continual Learning for Conditional Image Generation (2019, background): “Label-conditioned models [8, 18] have also been explored for generating images for specific categories.”
Good Semi-supervised Learning That Requires a Bad GAN (2017, positive): “Recently, Triple GAN [11] also employs the idea of conditional generator, but uses adversarial cost to match the two model-deﬁned factorizations of the joint distribution with the one deﬁned by paired data.”
A Bayesian Data Augmentation Approach for Learning Deep Models (2017, methodology): “Finally, our proposed GAN model is similar to the recently proposed triplet GAN [21] 1 , but it is important to emphasise that triplet GAN was proposed in order to improve the training procedure for GANs, while our model represents a particular realisation of the proposed Bayesian DA algorithm,…”
ALICE: Towards Understanding Adversarial Learning for Joint Distribution Matching (2017, background): “Conditional GAN and its variants [15, 18, 19, 20] have been widely used in supervised tasks.”
High-Fidelity Image Generation With Fewer Labels (2019, neutral): “Up to our knowledge, improvements in sample quality through partial label information are reported in Li et al. (2017); Deng et al. (2017); Sricharan et al. (2017), all of which consider only low-resolution data sets from a restricted domain.”
Self-Supervised Learning: Generative or Contrastive (2020, positive): “Based on BiGAN and ALI, later studies [20], [34] discover that GAN with deeper and larger networks and modified architectures can produce even better results on downstream task.”
Smooth Neighbors on Teacher Graphs for Semi-Supervised Learning (2017, methodology): “orm a tight and concentrated cluster. In (c) without SNTG, the cluster of the same class are divided into several parts. images with strange patterns. Some works have been done to analyze the reasons [15, 11, 26]. An interesting ﬁnding is that our method can also alleviate the problem. Fig. 6 presen”
Adversarial Symmetric Variational Autoencoder (2017, background): “To alleviate this problem, some researchers add a regularization term, such as reconstruction loss [24, 25, 26] or mutual information [4], to the GAN objective, to restrict the space of suitable mapping functions, thus avoiding some of the failure modes of GANs, i.”
Guided Collaborative Training for Pixel-wise Semi-Supervised Learning (2020, methodology): “The adversarial-based methods [11,25,39,43] assemble the discriminator from GAN [14], and try to match the latent distributions between labeled and unlabeled data through the image-level adversarial constraint.”
Small Sample Learning in Big Data Era (2018, methodology): “Typically, (Li et al., 2017a) achieved 5% error rate on SVHN dataset (Netzer et al., 2011) using 1000 examples (fewer than 1% of the whole samples).”
An Incremental Self-Labeling Strategy for Semi-Supervised Deep Learning Based on Generative Adversarial Networks (2020, background): “In [24], the authors present the triple generative adversarial net (triple-GAN), which consists of three players of a generator, a discriminator and a classifier, where the generator and classifier characterize the conditional distributions between images and labels, and the discriminator solely foc”
A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications (2020, background): “ive, and propose the deﬁnition of a preferred generator. A triangle GAN ( -GAN) [50] is proposed for semi-supervised cross-domain joint distribution matching and -GAN is closely related to Triple-GAN [51]. Madani et al. [232] used semi-supervised learning with GANs for chest X-ray classiﬁcation. Fut”
How Generative Adversarial Networks and Its Variants Work : An Overview of GAN (2017, background): “By estimating real data probability distribution pdata(x) where x exists in the d-dimensional real space R, generative models can be applied to various tasks such as semi-supervised learning [97], [22],[64], image translation [52],[56], [11], domain adaptation [2],[13],[100] and other academic domai”
KDGAN: Knowledge Distillation with Generative Adversarial Networks (2018, background): “Our KDGAN also consists of three players including two generators and a discriminator, but differs from Triple-GAN in that: (1) Both generators in KDGAN learn a conditional distribution over labels given features.”
Image Captioning with Very Scarce Supervised Data: Adversarial Semi-Supervised Learning Approach (2019, methodology): “Motivated by [Chongxuan et al., 2017], the formulation of this adversarial training is as follows:”
Semi-supervised Learning on Graphs with Generative Adversarial Nets (2018, background): “In addition, various auxiliary architectures are combined with semi-supervised GANs to classify images more accurately [5, 11, 21].”