Research map / Image generation / GANs
McGan: Mean and Covariance Feature Matching GAN
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 covariance feature matching instead of WGAN's mean matching | McGAN · CIFAR-10, conditional (+L) | inception score: 3.24 $\pm$ 0.02 -> 7.29 $\pm$ 0.06 | Helped |
| Use covariance feature matching instead of WGAN's mean matching | McGAN · CIFAR-10, unconditional (+L) | inception score: 5.21 $\pm$ 0.07 -> 6.97 $\pm$ 0.10 | Helped |
| Use covariance feature matching instead of WGAN's mean matching | McGAN · CIFAR-10, unconditional (-L) | inception score: 6.39 $\pm$ 0.07 -> 6.73 $\pm$ 0.04 | Helped |
| Add L1 mean matching to covariance matching (ablation) | McGAN · CIFAR-10, conditional (+L) | inception score: 7.29 $\pm$ 0.06 -> 7.11 $\pm$ 0.04 | Hurt |
| Add L1 mean matching to covariance matching (ablation) | McGAN · CIFAR-10, unconditional (+L) | inception score: 6.97 $\pm$ 0.10 -> 6.93 $\pm$ 0.07 | No clear effect |
| Add L1 mean matching to covariance matching (ablation) | McGAN · CIFAR-10, unconditional (-L) | inception score: 6.73 $\pm$ 0.04 -> 6.42 $\pm$ 0.09 | Hurt |
| Add L2 mean matching to covariance matching (ablation) | McGAN · CIFAR-10, conditional (+L) | inception score: 7.29 $\pm$ 0.06 -> 7.27 $\pm$ 0.04 | No clear effect |
| Add L2 mean matching to covariance matching (ablation) | McGAN · CIFAR-10, unconditional (+L) | inception score: 6.97 $\pm$ 0.10 -> 6.69 $\pm$ 0.08 | Hurt |
| Add L2 mean matching to covariance matching (ablation) | McGAN · CIFAR-10, unconditional (-L) | inception score: 6.73 $\pm$ 0.04 -> 6.35 $\pm$ 0.04 | Hurt |
| Vary the number of covariance components from 4 to 64 (ablation) | Covariance matching GAN · LFW and LSUN bedrooms | training stability: 4 to 64 | No clear effect |
Closest papers by meaning
- Sobolev GAN (ICLR 2018 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
- Sphere Generative Adversarial Network Based on Geometric Moment Matching (CVPR 2019 · similarity 0.86)
- Metropolis-Hastings Generative Adversarial Networks (ICML 2019 · similarity 0.86)
- SMaRt: Improving GANs with Score Matching Regularity (ICML 2024 · similarity 0.85)
- Bayesian GAN (NeurIPS 2017 · 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.