Research map / Image generation / GANs
MonoFlow: Rethinking Divergence GANs via the Perspective of Wasserstein Gradient Flows
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 logit loss instead of non-saturated loss (ablation) | GAN with discriminator d*(x)=log r(x) · MNIST test | FID: 4.3309 -> 4.4631 | Hurt |
| Use logit loss instead of non-saturated loss (ablation) | GAN with discriminator d*(x)=log r(x) · CIFAR-10 test | FID: 21.2980 -> 20.7969 | Helped |
| Use logit loss instead of non-saturated loss (ablation) | GAN with discriminator d*(x)=log r(x) · Celeb-A test | FID: 20.6109 -> 21.2240 | Hurt |
| Use Arcsin loss instead of non-saturated loss (ablation) | GAN with discriminator d*(x)=log r(x) · MNIST test | FID: 4.3309 -> 4.4893 | Hurt |
| Use Arcsin loss instead of non-saturated loss (ablation) | GAN with discriminator d*(x)=log r(x) · CIFAR-10 test | FID: 21.2980 -> 21.2533 | Helped |
| Use Arcsin loss instead of non-saturated loss (ablation) | GAN with discriminator d*(x)=log r(x) · Celeb-A test | FID: 20.6109 -> 21.0077 | Hurt |
| Use logit loss instead of vanilla loss (ablation) | GAN with discriminator d*(x)=r(x)/(1+r(x)) · MNIST test | FID: 5.0808 -> 4.6000 | Helped |
| Use logit loss instead of vanilla loss (ablation) | GAN with discriminator d*(x)=r(x)/(1+r(x)) · CIFAR-10 test | FID: 23.8330 -> 22.7969 | Helped |
| Use logit loss instead of vanilla loss (ablation) | GAN with discriminator d*(x)=r(x)/(1+r(x)) · Celeb-A test | FID: 21.6787 -> 20.5231 | Helped |
| Use Arcsin loss instead of vanilla loss (ablation) | GAN with discriminator d*(x)=r(x)/(1+r(x)) · MNIST test | FID: 5.0808 -> 4.5525 | Helped |
| Use Arcsin loss instead of vanilla loss (ablation) | GAN with discriminator d*(x)=r(x)/(1+r(x)) · CIFAR-10 test | FID: 23.8330 -> 23.4698 | Helped |
| Use Arcsin loss instead of vanilla loss (ablation) | GAN with discriminator d*(x)=r(x)/(1+r(x)) · Celeb-A test | FID: 21.6787 -> 22.1024 | Hurt |
Closest papers by meaning
- A Convex Duality Framework for GANs (NeurIPS 2018 · similarity 0.91)
- Flow-GAN: Combining Maximum Likelihood and Adversarial Learning in Generative Models (AAAI 2018 · similarity 0.90)
- Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Step (ICLR 2018 · similarity 0.89)
- MS$^3$D: A RG Flow-Based Regularization for GAN Training with Limited Data (ICML 2024 · similarity 0.88)
- Deep MMD Gradient Flow without adversarial training (ICLR 2025 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · 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.