Least Squares Generative Adversarial Networks

Mao, Xudong; Li, Qing; Xie, Haoran; Lau, Raymond Y. K.; Wang, Zhen; Smolley, Stephen Paul

doi:10.48550/arxiv.1611.04076

Cited by 107 publications

(87 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• several more recent GAN architectures that have been applied to the t t dataset: Wasserstein GAN (WGAN) [26], WGAN with Gradient Penalty (WGAN-GP) [27], Least Squares GAN (LS-GAN) [28], Maximum Mean Discrepancy GAN (MMDGAN) [29] and…”

Section: Methodsmentioning

confidence: 99%

“…where a,b and c are constants that need to be fixed and must satisfy b − a = 2 and b − c = 1 for effectively minimizing the Pearson χ 2 divergence [28]. By assuming an optimal discriminator D * , choosing a = −1, b = 1, c = 0 and adding a term 1 2 E x∼p d (D(x) − c) 2 to eq.…”

Section: B Generative Modelsmentioning

confidence: 99%

“…The LS-GAN does contain such a penalization such that G creates samples that are closer to the decision boundary and thereby also relieves the vanishing gradient problem. Even for simple problems the authors in [28] already show that while regular GANs have mode collapse, the LSGAN does not and recent advances in the application of generative models to physics problems also suggest that the LSGAN performs well for learning a probability density [39]. Our generator has 4 hidden layers with 512, 1024, 2048 and 2048 neurons, the LeakyReLU activation function [40] with a slope parameter α = 0.2 for the hidden layers and a hyperbolic tangent for the output layer and Batch normalization [41] with a momentum parameter of 0.8.…”

Section: B Generative Modelsmentioning

confidence: 99%

See 2 more Smart Citations

Event Generation and Statistical Sampling for Physics with Deep Generative Models and a Density Information Buffer

Otten¹,

Caron²,

Swart³

et al. 2019

Preprint

View full text Add to dashboard Cite

We present a study for the generation of events from a physical process with deep generative models. The simulation of physical processes requires not only the production of physical events, but also to ensure these events occur with the correct frequencies. We investigate the feasibility of learning the event generation and the frequency of occurrence with Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) to produce events like Monte Carlo generators. We study three processes: a simple two-body decay, the processes e + e − → Z → l + l − and pp → t t including the decay of the top quarks and a simulation of the detector response. We find that the tested GAN architectures and the standard VAE are not able to learn the distributions precisely. By buffering density information of encoded Monte Carlo events given the encoder of a VAE we are able to construct a prior for the sampling of new events from the decoder that yields distributions that are in very good agreement with real Monte Carlo events and are generated several orders of magnitude faster. Applications of this work include generic density estimation and sampling, targeted event generation via a principal component analysis of encoded ground truth data, anomaly detection and more efficient importance sampling, e.g. for the phase space integration of matrix elements in quantum field theories.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: B Generative Modelsmentioning

confidence: 99%

Section: B Generative Modelsmentioning

confidence: 99%

See 1 more Smart Citation

Event Generation and Statistical Sampling for Physics with Deep Generative Models and a Density Information Buffer

Otten¹,

Caron²,

Swart³

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Such a two-player game is actually optimized when they reach the Nash-Equilibrium point, that is, the discriminator can not tell whether an image is real or not. Many following works improve generative adversarial networks by better loss, training skills and evaluating metrics [1,13,19,34,16,24,27,35]. The objective of VAEs is to maximize the variational lower bound of log-likelihood of target data points.…”

Section: Three Basic Generative Modelsmentioning

confidence: 99%

“…Generative adversarial networks (GANs) [1,11,27,32] and variational auto-encoders (VAEs) [18] are two types of the most popular generative models due to their solid theoretic foundation and excellent results. Also, the performance of conditional image synthesis by these models improves rapidly with the fast development of deep learning.…”

Section: Introductionmentioning

confidence: 99%

Conditional Adversarial Generative Flow for Controllable Image Synthesis

Liu

Gong

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

Flow-based generative models show great potential in image synthesis due to its reversible pipeline and exact loglikelihood target, yet it suffers from weak ability for conditional image synthesis, especially for multi-label or unaware conditions. This is because the potential distribution of image conditions is hard to measure precisely from its latent variable z. In this paper, based on modeling a joint probabilistic density of an image and its conditions, we propose a novel flow-based generative model named conditional adversarial generative flow (CAGlow). Instead of disentangling attributes from latent space, we blaze a new trail for learning an encoder to estimate the mapping from condition space to latent space in an adversarial manner. Given a specific condition c, CAGlow can encode it to a sampled z, and then enable robust conditional image synthesis in complex situations like combining person identity with multiple attributes. The proposed CAGlow can be implemented in both supervised and unsupervised manners, thus can synthesize images with conditional information like categories, attributes, and even some unknown properties. Extensive experiments show that CAGlow ensures the independence of different conditions and outperforms regular Glow to a significant extent.

show abstract

Conditional generative adversarial networks based on the principle of homologycontinuity for face aging

Ning

Gou

Dong

et al. 2020

Concurrency and Computation

View full text Add to dashboard Cite

Age is one of the most important biological characteristics of the human face. The increase of age coincides with the increase of the aging degree of the face. Face aging synthesis is attracting increasingly more attention from domestic and overseas scholars in the computer vision and computer graphics fields, and it can be integrated into the basic research of face correlation, such as cross‐age face analysis and age estimation. At present, some achievements have been made in face aging synthesis research; however, it is still an urgent problem to reduce the number of parameters and computational complexity of the network while ensuring the aging effect. Therefore, a new face aging algorithm is proposed in this article. Unlike the previous methods of aging process simulation, we introduce an assisted age classification network based on the principle of homology continuity, which is more in line with the human cognition process. After pretraining, the result of age classification is improved, and the pretraining model is then added to the framework of aging face generation for fine‐tuning to constrain the generated aging face, which can improve the aging accuracy of the generated image. Furthermore, we reconstruct the input face by using the age tag of the input face and the synthesized aging face and maintain the identity invariance in the face aging process by minimizing the reconstruction loss. The experimental results show that the method proposed in this article produces a considerable effect of face aging and significantly reduces the number of parameters and the complexity of computational.

show abstract

Least Squares Generative Adversarial Networks

Cited by 107 publications

References 13 publications

Event Generation and Statistical Sampling for Physics with Deep Generative Models and a Density Information Buffer

Event Generation and Statistical Sampling for Physics with Deep Generative Models and a Density Information Buffer

Conditional Adversarial Generative Flow for Controllable Image Synthesis

Conditional generative adversarial networks based on the principle of homologycontinuity for face aging

Contact Info

Product

Resources

About