Research map / Label-efficient and robust learning / Recognition and long-tail
Combating Mode Collapse via Offline Manifold Entropy Estimation
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 MaEM-GAN instead of MaF-EwGAN | DCGAN · CIFAR-10 | FID: 30.85 -> 29.22 | Helped |
| Use MaEM-GAN instead of MaF-EwGAN | DCGAN · CelebA | FID: 12.43 -> 9.14 | Helped |
| Use MaEM-GAN instead of EBM-0GP | GAN image generation · ANIMEFACE | Inception Score: 2.26 -> 2.80 | Helped |
| Use MaEM-GAN instead of DDPM | Image generation · ANIMEFACE | FID: 8.81 -> 8.62 | Helped |
| Add MaEM-GAN learning objective | BigGAN · CIFAR-10 | FID: 16.04 -> 13.86 | Helped |
| Add MaEM-GAN learning objective | StyleGAN-V2 · FFHQ | FID: 2.84 -> 2.67 | Helped |
| Add DIsoMap to the high-dimensional WGAN baseline (ablation) | High-dimensional WGAN · CIFAR-10 | FID: >100 -> 51.96 | Helped |
| Add DLLE alongside DIsoMap (ablation) | High-dimensional WGAN · CIFAR-10 | FID: 51.96 -> 31.70 | Helped |
| Add RB-MaEM alongside DIsoMap (ablation) | High-dimensional WGAN · CIFAR-10 | FID: 51.96 -> 30.73 | Helped |
| Use adversarial noise rather than rotation and Gaussian noise for DLLE transformations (ablation) | MaEM-GAN · CIFAR-10 | FID: 29.22 -> 33.04 | Hurt |
| Set replay buffer size to 1024 rather than 128, 512, or 2048 (ablation) | MaEM-GAN · CIFAR-10 | FID: 31.58; 32.36; 29.22; 31.87 | Mixed |
| Add RB-MaEM (ablation) | MaEM-GAN · CIFAR-10 | I-Variance: 3.41 ± 0.03 -> 4.88 ± 0.04 | Helped |
Closest papers by meaning
- Dual Discriminator Generative Adversarial Nets (NeurIPS 2017 · similarity 0.89)
- Improving GAN with Neighbors Embedding and Gradient Matching (AAAI 2019 · similarity 0.89)
- UniGAN: Reducing Mode Collapse in GANs using a Uniform Generator (NeurIPS 2022 · similarity 0.89)
- PacGAN: The power of two samples in generative adversarial networks (NeurIPS 2018 · similarity 0.88)
- Structure-preserving GANs (ICML 2022 · similarity 0.88)
- VEEGAN: Reducing Mode Collapse in GANs using Implicit Variational Learning (NeurIPS 2017 · 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.