Research map / Image generation / GANs
Improving GANs Using Optimal Transport
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 optimal transport matching within mini-batch energy distance (vs random matching) (ablation) | OT-GAN on CIFAR-10 · CIFAR-10 | Inception score: 8.47 vs 4.64 | Helped |
| learn transport cost adversarially in critic feature space (vs fixed cosine on pixels) (ablation) | OT-GAN on CIFAR-10 · CIFAR-10 | Inception score: 8.47 vs 4.93 | Helped |
| increase mini-batch size (200 to 8000) (ablation) | OT-GAN on CIFAR-10 · CIFAR-10 | Inception score | Helped |
| OT-GAN objective with mini-batch energy distance vs DAN-S loss under fixed discriminator (consistency test) | GAN on 8-Gaussian mixture · 8 Gaussians toy | modes covered after fixing critic: covers all 8 modes vs collapse to 3 | Helped |
| OT-GAN instead of DCGAN variant on ImageNet dogs | OT-GAN · ImageNet Dogs 128x128 | Inception score: 8.97 vs 8.19 | Helped |
| update generator more often than critic (3:1) | OT-GAN on CIFAR-10 · CIFAR-10 | qualitative stability | No clear effect |
Closest papers by meaning
- Parallel Optimal Transport GAN (CVPR 2019 · similarity 0.93)
- Analyzing and Improving Optimal-Transport-based Adversarial Networks (ICLR 2024 · similarity 0.91)
- Discriminator optimal transport (NeurIPS 2019 · similarity 0.91)
- On the Convergence and Robustness of Training GANs with Regularized Optimal Transport (NeurIPS 2018 · similarity 0.90)
- Projected GANs Converge Faster (NeurIPS 2021 · similarity 0.90)
- Generative Modeling with Optimal Transport Maps (ICLR 2022 · similarity 0.89)
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.