Wasserstein Generative Adversarial Networks

By Martín Arjovsky, Soumith Chintala, L. Eon Bottou. 2017

9,668 citations recorded by the source corpus.

How researchers see this paper

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

In Generative Models

Citations in the corpus

UNITER: Learning UNiversal Image-TExt Representations (2019, methodology): “ high-dimensional and continuous, thus cannot be supervised via class likelihood. Instead, we propose three variants for MRM, which share the same objective base: L MRM() = E (w;v)˘Df (v mjv nm;w): (4) 1) Masked Region Feature Regression (MRFR)MRFR learns to regress the Transformer output of each ma”
CTRL: A Conditional Transformer Language Model for Controllable Generation (2019, methodology): “Much research then focused on methods for controlling the generation process and improving estimation of generative distributions (Arjovsky et al., 2017; Chen et al., 2016; Kingma & Welling, 2013).”
Are GANs Created Equal? A Large-Scale Study (2017, background): “In Wasserstein GAN [1] the discriminator is allowed to output a real number and the objective function is equivalent to the MM GAN loss without the sigmoid (WGAN in Table 1).”
Generative Modeling by Estimating Gradients of the Data Distribution (2019, methodology): “The former uses log-likelihood (or a suitable surrogate) as the training objective, while the latter uses adversarial training to minimize f -divergences [40] or integral probability metrics [2, 55] between model and data distributions.”
On the Anatomy of MCMC-based Maximum Likelihood Learning of Energy-Based Models (2019, methodology): “The W-GAN [1] framework is adapted to the INN method in the WINN model [22].”
Large Scale GAN Training for High Fidelity Natural Image Synthesis (2018, background): “First, one might consider an unbounded loss (such as the Wasserstein loss (Arjovsky et al., 2017)) which would not suffer this gradient attentuation.”
Autoregressive Quantile Networks for Generative Modeling (2018, methodology): “Recently, Arjovsky et al. (2017) introduced the Wasserstein GAN, reposing the two-player game as the estimation of the gradient of the 1 -Wasserstein distance between the data and generator distributions.”
Spectral Normalization for Generative Adversarial Networks (2018, methodology): “We also compare the performance quality of the algorithm against those of other regularization/normalization techniques for the discriminator networks, including: Weight clipping (Arjovsky et al., 2017), WGAN-GP (Gulrajani et al., 2017), batch-normalization (BN) (Ioffe & Szegedy, 2015), layer normal”
Generative Adversarial Networks (2021, positive): “For example, the Wasserstein GAN (WGAN) [7] replaces the original GAN Loss Function with the Wasserstein distance, which can produce more stable training.”
Dual Contradistinctive Generative Autoencoder (2020, neutral): “The subsequent development after GAN [55, 2, 17, 34, 15, 14, 13] has led to a great leap forward in building decoder-based generative models.”
Learning Layout and Style Reconfigurable GANs for Controllable Image Synthesis (2020, background): “GANs [1] are able to synthesize realistic and high resolution images under various settings, including both unconditional [2], [7], [8], [34] and conditional tasks [5], [6], [9].”
A U-Net Based Discriminator for Generative Adversarial Networks (2020, background): “Note that we do not modify the generator in any way, and our work is orthogonal to the ongoing research on architectural changes of the generator [20, 27], divergence measures [25, 1, 37], and regularizations [40, 15, 34].”
Imperfect ImaGANation: Implications of GANs Exacerbating Biases on Facial Data Augmentation and Snapchat Selfie Lenses (2020, methodology): “We show experiments on two other GAN architectures designed to address mode collapse – Wasserstein GAN (Arjovsky et al., 2017) and AdaGAN (Tolstikhin et al., 2017) – in the appendix.”
Semantic Image Synthesis With Spatially-Adaptive Normalization (2019, methodology): “s function used in pix2pixHD except that we replace the least squared loss term [28] with the hinge loss term [25,30,45]. We test several ResNet-based discriminators used in recent unconditional GANs [1,29,31] but observe similar results at the cost of a higher GPU memory requirement. Adding the SPA”
Diagnosing and Enhancing VAE Models (2019, background): “7 Additionally, we present results from (Lucic et al., 2018) involving numerous competing GAN models, including MM GAN (Goodfellow et al., 2014), WGAN (Arjovsky et al., 2017), WGAN-GP (Gulrajani et al., 2017), NS GAN (Fedus et al., 2017), DRAGAN (Ko-dali et al., 2017), LS GAN (Mao et al., 2017) and ”
Cross-Modal Contrastive Learning for Text-to-Image Generation (2021, background): “A large amount of work has focused on designing the adversarial objective to improve training [12, 1, 31, 47, 29, 54].”
Aligning Language Models with Preferences through f-divergence Minimization (2023, background): “Theoretical insight on the relationship between divergence choice and the convergence of probability distributions was provided by Arjovsky et al. (2017).”
Stochastic Hamiltonian Gradient Methods for Smooth Games (2020, neutral): “For WGAN, we see that stochastic SGDA fails to converge and that L-SVRHG is the only method to converge linearly on the Hamiltonian.”
A Closer Look at the Optimization Landscapes of Generative Adversarial Networks (2019, methodology): “On the contrary, WGAN-GP has eigenvalues with very large imaginary parts and relatively small real eigenvalues (Fig.”