HGAN: Hyperbolic Generative Adversarial Network
Recently, Hyperbolic Spaces in the context of Non-Euclidean Deep Learning have gained popularity because of their ability to represent hierarchical data. We propose that it is possible to take advantage of the hierarchical characteristic present in the images by using hyperbolic neural networks in a...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2021-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9474500/ |
| _version_ | 1811274076267741184 |
|---|---|
| author | Diego Lazcano Nicolas Fredes Franco Werner Creixell |
| author_facet | Diego Lazcano Nicolas Fredes Franco Werner Creixell |
| author_sort | Diego Lazcano |
| collection | DOAJ |
| description | Recently, Hyperbolic Spaces in the context of Non-Euclidean Deep Learning have gained popularity because of their ability to represent hierarchical data. We propose that it is possible to take advantage of the hierarchical characteristic present in the images by using hyperbolic neural networks in a GAN architecture. In this study, different configurations using fully connected hyperbolic layers in the GAN, WGAN, CGAN, and the mapping network of the StyleGAN2 are tested in what we call the HGAN, HWGAN, HCGAN, and HStyleGAN, respectively. Furthermore, we test multiple values of curvature and introduce an exponential way to train it. The results are measured using the Inception Score (IS) and the Fréchet Inception Distance (FID) over the MNIST dataset and with FID over CIFAR-10. Depending on the configuration and space curvature, better results are achieved for each proposed hyperbolic version than their euclidean counterpart. |
| first_indexed | 2024-04-12T23:12:25Z |
| format | Article |
| id | doaj.art-1aecc71fa4e449b189523f416dbc07b2 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-12T23:12:25Z |
| publishDate | 2021-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-1aecc71fa4e449b189523f416dbc07b22022-12-22T03:12:46ZengIEEEIEEE Access2169-35362021-01-019963099632010.1109/ACCESS.2021.30947239474500HGAN: Hyperbolic Generative Adversarial NetworkDiego Lazcano0Nicolas Fredes Franco1Werner Creixell2https://orcid.org/0000-0002-6647-6429Departamento de Ingeniería Electrónica, Universidad Técnica Federico Santa María, Valparaíso, ChileDepartamento de Ingeniería Electrónica, Universidad Técnica Federico Santa María, Valparaíso, ChileDepartamento de Ingeniería Electrónica, Universidad Técnica Federico Santa María, Valparaíso, ChileRecently, Hyperbolic Spaces in the context of Non-Euclidean Deep Learning have gained popularity because of their ability to represent hierarchical data. We propose that it is possible to take advantage of the hierarchical characteristic present in the images by using hyperbolic neural networks in a GAN architecture. In this study, different configurations using fully connected hyperbolic layers in the GAN, WGAN, CGAN, and the mapping network of the StyleGAN2 are tested in what we call the HGAN, HWGAN, HCGAN, and HStyleGAN, respectively. Furthermore, we test multiple values of curvature and introduce an exponential way to train it. The results are measured using the Inception Score (IS) and the Fréchet Inception Distance (FID) over the MNIST dataset and with FID over CIFAR-10. Depending on the configuration and space curvature, better results are achieved for each proposed hyperbolic version than their euclidean counterpart.https://ieeexplore.ieee.org/document/9474500/GANWGANCGANStyleGAN2hyperbolic spacesPoincaré ball |
| spellingShingle | Diego Lazcano Nicolas Fredes Franco Werner Creixell HGAN: Hyperbolic Generative Adversarial Network IEEE Access GAN WGAN CGAN StyleGAN2 hyperbolic spaces Poincaré ball |
| title | HGAN: Hyperbolic Generative Adversarial Network |
| title_full | HGAN: Hyperbolic Generative Adversarial Network |
| title_fullStr | HGAN: Hyperbolic Generative Adversarial Network |
| title_full_unstemmed | HGAN: Hyperbolic Generative Adversarial Network |
| title_short | HGAN: Hyperbolic Generative Adversarial Network |
| title_sort | hgan hyperbolic generative adversarial network |
| topic | GAN WGAN CGAN StyleGAN2 hyperbolic spaces Poincaré ball |
| url | https://ieeexplore.ieee.org/document/9474500/ |
| work_keys_str_mv | AT diegolazcano hganhyperbolicgenerativeadversarialnetwork AT nicolasfredesfranco hganhyperbolicgenerativeadversarialnetwork AT wernercreixell hganhyperbolicgenerativeadversarialnetwork |