Research map / Image generation / GANs
MMD GAN: Towards Deeper Understanding of Moment Matching Network
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 an adversarially learned kernel instead of a fixed Gaussian kernel | GMMN · CIFAR-10 | Inception score: 3.47 ± .03 -> 6.17 ± .07 | Helped |
| Use an adversarially learned kernel instead of a fixed feature transformation | GMMN-C · CIFAR-10 | Inception score: 3.94 ± .04 -> 6.17 ± .07 | Helped |
| Use adversarial kernel learning rather than first-order moment matching | WGAN · CIFAR-10 | Inception score: 5.88 ± .07 -> 6.17 ± .07 | Helped |
| Train with batch size 64 instead of batch size 1024 | GMMN · CIFAR-10 | generated image quality | Hurt |
| Reduce the discriminator feasible set using the mean-feature constraint (ablation) | MMD GAN · Training | training speed and stability | Helped |
| Replace weight clipping with gradient penalty and omit the autoencoder reconstruction loss (ablation) | MMD GAN · CIFAR-10 and CelebA | generated image quality | No clear effect |
| Omit the autoencoder reconstruction loss (ablation) | MMD GAN · GAN training | training success | No clear effect |
Closest papers by meaning
- Improving Generative Moment Matching Networks with Distribution Partition (AAAI 2021 · similarity 0.90)
- Conditional Generative Moment-Matching Networks (NeurIPS 2016 · similarity 0.89)
- Demystifying MMD GANs (ICLR 2018 · similarity 0.87)
- Metropolis-Hastings Generative Adversarial Networks (ICML 2019 · similarity 0.86)
- Exact Gaussian Moment Matching for Residual Networks: a Second-Order Method (NeurIPS 2026 · similarity 0.85)
- SMaRt: Improving GANs with Score Matching Regularity (ICML 2024 · similarity 0.85)
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.