Research map / Image generation / GANs
Dynamically Grown Generative Adversarial Networks
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 |
|---|---|---|---|
| Dynamically search generator and discriminator growth while progressively training inherited architectures | GAN image generation · CIFAR-10 | FID: 18.33 -> 12.10 | Helped |
| Dynamically search generator and discriminator growth while progressively training inherited architectures | GAN image generation · LSUN 256×256 | FID: 10.76 -> 8.22 | Helped |
| Dynamically search generator and discriminator growth while progressively training inherited architectures | GAN image generation · CIFAR-10 | Inception Score: 8.80 ± 0.05 -> 8.64 ± 0.06 | Mixed |
| Allow generator and discriminator parameter counts to be asymmetric | DGGAN · CIFAR-10 | normalized FID | Helped |
| Use smaller rather than larger discriminator convolution filters | DGGAN discriminator · CIFAR-10 search candidates | normalized FID improvement | Helped |
| Vary generator convolution filter size rather than fixing one size | DGGAN generator · CIFAR-10 search candidates | normalized FID improvement | No clear effect |
| Add more convolution filters | DGGAN generator and discriminator · CIFAR-10 search candidates | normalized FID improvement | Mixed |
| Increase convolution-layer complexity | DGGAN generator · CIFAR-10 search candidates | normalized FID improvement | Hurt |
| Prune sub-optimal parent candidates and expand top-performing candidates | DGGAN search · CIFAR-10 search | likelihood of generating good children | No clear effect |
| Reduce top-K search width from 20 to 8 | DGGAN search · CIFAR-10 simulation | generation quality | No clear effect |
| Reduce random action sampling ratio from 0.4 to 0.2 | DGGAN search · CIFAR-10 simulation | generation quality | No clear effect |
| Apply the searched AutoGAN architecture at higher resolution | AutoGAN architecture · LSUN 64×64 | FID: 19.74 -> 19.61 | Hurt |
Closest papers by meaning
- Progressive Growing of GANs for Improved Quality, Stability, and Variation (ICLR 2018 · similarity 0.92)
- Towards Principled Methods for Training Generative Adversarial Networks (ICLR 2017 · similarity 0.92)
- Improving GANs with A Dynamic Discriminator (NeurIPS 2022 · similarity 0.91)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.90)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · similarity 0.90)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.89)
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.