Research map / Image generation / GANs
A Characteristic Function Approach to Deep Implicit Generative Modeling
Check 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 |
|---|---|---|---|
| Optimize the weighting-distribution scale instead of fixing it (ablation) | CF-GAN with Gaussian weighting · CIFAR10 | FID: 39.81 (0.93) -> 38.47 (1.00) | Helped |
| Add gradient penalty (ablation) | OCF-GAN with Gaussian weighting · CIFAR10 | FID: 38.47 (1.00) -> 33.08 (0.26) | Helped |
| Add gradient penalty (ablation) | OCF-GAN with Gaussian weighting · STL10 | FID: 32.51 (0.87) -> 26.16 (0.64) | Helped |
| Add gradient penalty (ablation) | OCF-GAN with Gaussian weighting · CelebA | FID: 14.91 (0.83) -> 9.39 (0.25) | Helped |
| Increase sampled frequencies from 1 to 16 (ablation) | OCF-GAN-GP · MNIST | FID: 0.44 (0.03) -> 0.35 (0.02) | Helped |
| Increase sampled frequencies from 16 to 64 (ablation) | OCF-GAN-GP · MNIST | FID: 0.35 (0.02) -> 0.36 (0.07) | No clear effect |
| Remove the decoder network (ablation) | CF-GAN · Image generation benchmarks | model performance | No clear effect |
| Use Student-t rather than Gaussian weighting (ablation) | OCF-GAN-GP · CIFAR10 | FID: 33.08 (0.26) -> 34.33 (0.77) | Hurt |
| Optimize weighting-distribution parameters to increase two-sample test power (ablation) | ECFD and smoothed ECFD · High-dimensional Gaussian two-sample test | test power | Helped |
| Optimize individual frequencies directly rather than the weighting distribution (ablation) | ECFD two-sample test · High-dimensional Gaussian two-sample test | null-hypothesis acceptance | Hurt |
Closest papers by meaning
- Reciprocal Adversarial Learning via Characteristic Functions (NeurIPS 2020 · similarity 0.89)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.88)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.88)
- A Geometric Analysis of Deep Generative Image Models and Its Applications (ICLR 2021 · similarity 0.87)
- Consistency-GAN: Training GANs with Consistency Model (AAAI 2024 · similarity 0.86)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · 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.