Open Questions about Generative Adversarial Networks

By Augustus Odena. 2019

37 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Consistency Regularization for Generative Adversarial Networks (2019, background): “In fact, formally characterizing the convergence properties of the GAN training procedure is mostly an open problem (Odena, 2019).”
Bias Correction of Learned Generative Models using Likelihood-Free Importance Weighting (2019, background): “For example, images generated by deep generative models often possess distinct artifacts which can be exploited by the classifier to give highly-confident predictions [23, 24].”
Mimicry: Towards the Reproducibility of GAN Research (2020, background): “Extraordinary progress has been made in the ﬁeld of generative modelling following the introduction of Generative Adversarial Networks (GANs) [11], a form of generative model known for its sample efﬁciency in generating high resolution data [31].”
Improved Consistency Regularization for GANs (2020, background): “There is so much related work on GANs (Goodfellow et al., 2014) that it is impossible to do it justice (see Odena (2019); Kurach et al. (2019) for different overviews of the ﬁeld), 2020) and correspondingly report better scores).”
ODE Analysis of Stochastic Gradient Methods with Optimism and Anchoring for Minimax Problems and GANs (2019, unknown): “Despite the remarkable empirical success of GANs, further understanding the global training dynamics empirically and theoretically is considered a major open problem [18, 54, 39, 37, 48].”
Adversarial Text-to-Image Synthesis: A Review (2021, background): “Other Generative Methods Current T2I methods are heavily based on GANs which still have many open problems despite the remarkable progress during the last few years [146].”
Pros and Cons of GAN Evaluation Measures: New Developments (2021, background): “IS does not capture intra-class diversity, is insensitive to the prior distribution over labels (hence is biased towards ImageNet dataset and Inception model 4 ), and is very sensitive to model parameters and implementations (Barratt and Sharma, 2018; Odena, 2019).”