Research map / Image generation / GANs
Reciprocal Adversarial Learning via Characteristic Functions
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 |
|---|---|---|---|
| remove the reciprocal loss (lambda=0) (ablation) | RCF-GAN · CelebA generation | FID: 59.39 -> 15.86 | Helped |
| remove the reciprocal loss (lambda=0) (ablation) | RCF-GAN · CelebA reconstruction | FID: 100 -> 14.82 | Helped |
| remove the anchor design (ablation) | RCF-GAN · CelebA generation | FID: 17.80 -> 15.86 | Helped |
| remove the anchor design (ablation) | RCF-GAN · CelebA reconstruction | FID: 100 -> 14.82 | Helped |
| use t-net instead of fixed Gaussian samples for t (ablation) | RCF-GAN · CelebA | FID: 19.34 -> 15.86 | Helped |
| use t-net instead of fixed Gaussian samples for t (ablation) | RCF-GAN · LSUN Bedroom | FID: 38.16 -> 40.15 | Hurt |
| use t-net instead of fixed Gaussian samples for t (ablation) | RCF-GAN · CIFAR-10 | FID: 31.55 -> 31.21 | Helped |
| train generator only via the CF loss without critic latent comparison (ablation) | RCF-GAN · CelebA | FID: 165 FID | Hurt |
| set alpha to 0.75 (more amplitude) (ablation) | RCF-GAN · CelebA | FID: 15.86 -> 13.84 | Helped |
| set alpha to 0.75 (more amplitude) (ablation) | RCF-GAN · CelebA | KID: 0.011 -> 0.009 | Helped |
| vary alpha from 0.1 to 0.9 (ablation) | RCF-GAN · CelebA/CIFAR-10 | FID | No clear effect |
| add reciprocal auto-encoder structure with CF loss, versus OCF-GAN-GP | RCF-GAN · CIFAR-10 | FID: 33.68 -> 31.21 | Helped |
Closest papers by meaning
- A Characteristic Function Approach to Deep Implicit Generative Modeling (CVPR 2020 · similarity 0.89)
- Quantitatively Evaluating GANs With Divergences Proposed for Training (ICLR 2018 · similarity 0.87)
- Approximation and Convergence Properties of Generative Adversarial Learning (NeurIPS 2017 · similarity 0.87)
- PCF-GAN: generating sequential data via the characteristic function of measures on the path space (NeurIPS 2023 · similarity 0.86)
- Structure-preserving GANs (ICML 2022 · similarity 0.86)
- A Convex Duality Framework for GANs (NeurIPS 2018 · similarity 0.86)
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.