Improving Generative Adversarial Networks with Denoising Feature Matching

By David Warde-Farley, Yoshua Bengio. 2016

178 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

Learning Descriptor Networks for 3D Shape Synthesis and Analysis (2018, methodology): “To quantitatively evaluate our model, we adopt the Inception score proposed by [26], which uses a reference convolutional neural network to compute where c denotes category, { ˜ Y i , i = 1 ,..., ˜ n } are synthesized examples sampled from the model, p ( c | ˜ Y ) is obtained from the output of the…”
Improved Training of Wasserstein GANs (2017, methodology): “Unsupervised Method Score ALI [8] (in [26]) 5.”
Stacked Generative Adversarial Networks (2016, background): “Serge Belongie is supported in part by a Google Focused Research Award.”
InfoMax-GAN: Improved Adversarial Image Generation via Information Maximization and Contrastive Learning (2020, methodology): “These sets of hyperparameters are based on settings in [3], [12], [29], [46], [47], which are well-tested settings used in popular works such as SNGAN and WGAN-GP [3].”
MMD GAN: Towards Deeper Understanding of Moment Matching Network (2017, background): “Although both WGAN and MMD-GAN can generate sharp images as we show in Section 5.1, our score is better than other GAN techniques except for DFM [32].”
HYPE: A Benchmark for Human eYe Perceptual Evaluation of Generative Models (2019, methodology): “We follow existing work on evaluating generative output by sampling K = 5000 generated images from each model [52, 41, 58] and K = 5000 real images from the training set.”
Variational Inference using Implicit Distributions (2017, methodology): “Sønderby et al. (2017) and Warde-Farley & Bengio (2017) demonstrated an alternative or complementary technique based on de-noising autoencoders.”
Multi-Generator Generative Adversarial Nets (2017, methodology): “…± 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 – – DCGAN (Radford et al., 2015) 6.40 ± 0.05 7.54 7.89 DFM (Warde-Farley & Bengio, 2016) 7 open less and less often.”
Dual Discriminator Generative Adversarial Nets (2017, methodology): “Alternatively, around the same time, there are various attempts to employ autoencoders as regularizers or auxiliary losses to penalize missing modes [5, 30, 4, 29].”
Memorization Precedes Generation: Learning Unsupervised GANs with Memory Networks (2018, unknown): “…et al., 2016) 6.86 ± 0.06 GAN + historical averaging - + minibatch discrimination EGAN-Ent-VI (Dai et al., 2017) 7.07 ± 0.10 Energy based Decoder net DFM (Warde-Farley & Bengio, 2017) 7.72 ± 0.13 Energy based Decoder net WGAN-GP (Gulrajani et al., 2017) 7.86 ± 0.07 Wasserstein + gradient penalty -…”