Investigating Shifts in GAN Output-Distributions
- A fundamental and still largely unsolved question in the context of Generative Adversarial Networks is whether they are truly able to capture the real data distribution and, consequently, to sample from it. In particular, the multidimensional nature of image distributions leads to a complex evaluation of the diversity of GAN distributions. Existing approaches provide only a partial understandingA fundamental and still largely unsolved question in the context of Generative Adversarial Networks is whether they are truly able to capture the real data distribution and, consequently, to sample from it. In particular, the multidimensional nature of image distributions leads to a complex evaluation of the diversity of GAN distributions. Existing approaches provide only a partial understanding of this issue, leaving the question unanswered. In this work, we introduce a loop-training scheme for the systematic investigation of observable shifts between the distributions of real training data and GAN generated data. Additionally, we introduce several bounded measures for distribution shifts, which are both easy to compute and to interpret. Overall, the combination of these methods allows an explorative investigation of innate limitations of current GAN algorithms. Our experiments on different data-sets and multiple state-of-the-art GAN architectures show large shifts between input and output distributions, showing that existing theoretical guarantees towards the convergence of output distributions appear not to be holding in practice.…
Document Type: | Conference Proceeding |
---|---|
Conference Type: | Konferenzartikel |
Zitierlink: | https://opus.hs-offenburg.de/5297 | Bibliografische Angaben |
Title (English): | Investigating Shifts in GAN Output-Distributions |
Conference: | Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021), Sydney, Australia |
Author: | Ricard Durall Lopez, Janis KeuperStaff MemberORCiDGND |
Year of Publication: | 2021 |
Page Number: | 8 |
First Page: | 1 |
Last Page: | 8 |
Parent Title (English): | [35th Conference on Neural Information Processing Systems] |
URL: | https://openreview.net/pdf?id=HPOZLHaMxQo |
Language: | English | Inhaltliche Informationen |
Institutes: | Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) |
Forschung / IMLA - Institute for Machine Learning and Analytics | |
Institutes: | Bibliografie |
Tag: | Computer Vision; Machine Learning; Pattern Recognition | Formale Angaben |
Open Access: | Open Access |
Licence (German): | Creative Commons - CC0 1.0 - Universell - Public Domain Dedication |
ArXiv Id: | http://arxiv.org/abs/2112.14061 |