Research map / Label-efficient and robust learning / Domain adaptation and generalization
Multi-marginal Wasserstein GAN
Jiezhang Cao, Langyuan Mo, Yifan Zhang, Kui Jia, Chunhua Shen, Mingkui Tan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use MWGAN's shared-potential multi-marginal objective instead of the strongest baseline | CelebA single-attribute translation · CelebA hair and pale skin | FID: 20.45 -> 19.63; 18.09 -> 15.91 | Improved |
| Use MWGAN's shared-potential multi-marginal objective instead of the strongest baseline | CelebA single-attribute translation · CelebA eyeglasses and mustache | FID: 23.69 -> 22.94; 23.75 -> 23.69 | Improved |
| Use MWGAN instead of the strongest baseline | CelebA single-attribute translation · CelebA hair, eyeglasses, mustache, and pale skin | classification accuracy: 96.00 -> 97.65; 99.51 -> 99.53; 99.06 -> 98.35; 92.48 -> 93.66 | Mixed |
| Use MWGAN instead of the strongest baseline | CelebA multi-attribute synthesis · B+E, B+M, B+M+E, and B+M+E+P | classification accuracy: 72.53 -> 75.82; 62.20 -> 69.01; 45.77 -> 53.75; 8.54 -> 19.95 | Improved |
| Use MWGAN instead of the strongest baseline | Edge→CelebA translation · Black, blond, and brown hair | FID: 53.41 -> 33.81; 81.00 -> 51.87; 57.51 -> 35.24 | Improved |
| Use a jointly matched multi-marginal objective rather than pairwise or mixed-distribution matching | Toy distribution matching with dissimilar source and target distributions · Toy Gaussian source to Gaussian or Uniform targets | distribution matching ability | Improved |
| Apply MWGAN to highly imbalanced domains instead of the compared translation methods | Style painting translation · Real scene to Monet, Van Gogh, and Ukiyo-e | visual quality | Improved |
Closest papers by meaning
- Improving the Improved Training of Wasserstein GANs: A Consistency Term and Its Dual Effect (ICLR 2018 · similarity 0.89)
- Wasserstein Divergence for GANs (ECCV 2018 · similarity 0.89)
- Some Theoretical Insights into Wasserstein GANs (JMLR 2021 · similarity 0.88)
- A Two-Step Computation of the Exact GAN Wasserstein Distance (ICML 2018 · similarity 0.88)
- Don't let your Discriminator be fooled (ICLR 2019 · similarity 0.88)
- Risk Bounds for Unsupervised Cross-Domain Mapping with IPMs (JMLR 2021 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
