Supplementary Materials for : “ f-GAN : Training Generative Neural Samplers using Variational Divergence Minimization ”

By Sebastian Nowozin, Botond Cseke. 2016

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

Aligning Language Models with Preferences through f-divergence Minimization (2023, background): “Nowozin et al. (2016) proposed a variational expression of f -divergences as a loss function for GANs.”
cGANs with Projection Discriminator (2018, neutral): “The most distinctive feature of GANs is the discriminator D ( x ) that evaluates the divergence between the current generative distribution p G ( x ) and the target distribution q ( x ) (Goodfellow et al., 2014; Nowozin et al., 2016; Arjovsky et al., 2017).”
A Variational Inequality Perspective on Generative Adversarial Networks (2018, neutral): “Much recent work has tried to tackle these training difﬁculties, usually by proposing new formulations of the GAN objective (Nowozin et al., 2016; Arjovsky et al., 2017).”
Disentangling by Factorising (2018, neutral): “This penalty is expressed as a KL divergence between this marginal distribution and the product of its marginals, and is optimised using a discriminator network following the divergence minimisation view of GANs (Nowozin et al., 2016; Mohamed & Lakshminarayanan, 2016).”
GradientDICE: Rethinking Generalized Offline Estimation of Stationary Values (2020, background): “…a probability measure, Zhang et al. (2020) propose to compute τ ∗ by solving the following optimization problem: where τ 0 is elementwise greater or equal, D φ is an f - divergence (Nowozin et al., 2016) associated with a convex, lower semi-continuous generator function φ : R + → R with φ (1) = 0 .”
On Variational Bounds of Mutual Information (2019, neutral): “Letting a ( y ) be the constant e recovers the bound of Nguyen, Wainwright, and Jordan (Nguyen et al., 2010) also known as f -GAN KL (Nowozin et al., 2016) and MINE-f (Belghazi et al., 2018) 1 : This tractable bound no longer requires learning a ( y ) , but now f ( x, y ) must learn to…”
Adversarial Feature Matching for Text Generation (2017, methodology): “Quantitative comparison We evaluate the generated-sentence quality using the BLEU score (Papineni et al., 2002) and Kernel Density Estimation (KDE), as in Good-fellow et al. (2014); Nowozin et al. (2016).”
LOGAN: Latent Optimisation for Generative Adversarial Networks (2019, background): “This procedure can be summarised as the following min-max game: The exact form of h ( · ) depends on the choice of loss function (Goodfellow et al., 2014; Arjovsky et al., 2017; Nowozin et al., 2016).”
Generative Adversarial Networks for Extreme Learned Image Compression (2018, neutral): “Furthermore, it uses a non-standard GAN formulation that does not (to the best of our knowledge) have an interpretation in terms of divergences between probability distributions, as in (Goodfellow et al., 2014; Nowozin et al., 2016).”
Maximum Entropy Generators for Energy-Based Models (2019, methodology): “In this paper, we used the Deep INFOMAX (DIM) estimator (Hjelm et al., 2018), which is based on maximizing the Jensen-Shannon divergence between the joint and the marginal (see (Nowozin et al., 2016) for the original F-GAN formulation). where s + ( a ) = log(1 + e a ) is the softplus function.”
PA-GAN: Improving GAN Training by Progressive Augmentation (2018, methodology): “Another line of work resorts to objective function re-formulation, e.g. by using the Pearson χ 2 divergence (Mao et al., 2016), the Wasserstein distance (Arjovsky et al., 2017), or f-divergence (Nowozin et al., 2016).”