Research map / Optimization / Optimizers
A Variational Inequality Perspective on 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 |
|---|---|---|---|
| Add an extrapolation step to Adam | GAN training · CIFAR-10 WGAN (DCGAN) | inception score: 6.05 ± .12 -> 6.38± .09 | Helped |
| Add an extrapolation step and uniform averaging to Adam | GAN training · CIFAR-10 WGAN-GP (ResNet) | inception score: 7.51 ± .17 -> 8.47 ± .10 | Helped |
| Add uniform averaging to ExtraAdam (ablation) | ExtraAdam · CIFAR-10 WGAN-GP (ResNet) | inception score: 7.90 ± .11 -> 8.47 ± .10 | Helped |
| Add uniform averaging to ExtraAdam (ablation) | ExtraAdam · CIFAR-10 WGAN (DCGAN) | inception score: 6.38± .09 -> 6.38 ± .20 | No clear effect |
| Use exponential moving average with ExtraAdam instead of no averaging (ablation) | ExtraAdam · CIFAR-10 WGAN (DCGAN) | inception score: 6.38± .09 -> 6.37 ± .08 | No clear effect |
| Use exponential moving average with ExtraAdam instead of no averaging (ablation) | ExtraAdam · CIFAR-10 WGAN-GP (ResNet) | inception score: 7.90 ± .11 -> 8.13 ± .07 | Helped |
| Use extrapolation from the past instead of standard extrapolation (ablation) | Adam · CIFAR-10 WGAN (DCGAN), no averaging | inception score: 6.38± .09 -> 5.98 ± .15 | Hurt |
| Use extrapolation from the past instead of standard extrapolation (ablation) | Adam · CIFAR-10 WGAN-GP (ResNet), no averaging | inception score: 7.90 ± .11 -> 7.84 ± .06 | No clear effect |
| Add extrapolation to Adam | GAN training · CIFAR-10 WGAN-GP (ResNet), no averaging | inception score: 7.98 ± .08 -> 7.90 ± .11 | No clear effect |
| Use extrapolation with uniform averaging rather than Optimistic Adam with uniform averaging | Adam · CIFAR-10 WGAN-GP (ResNet) | inception score: 8.18 ± .09 -> 8.47 ± .10 | Helped |
| Use extrapolation from the past rather than standard extrapolation (ablation) | Adam with uniform averaging · CIFAR-10 WGAN-GP (ResNet) | inception score: 8.47 ± .10 -> 8.01± .09 | Hurt |
Closest papers by meaning
- Variational Annealing of GANs: A Langevin Perspective (ICML 2019 · similarity 0.92)
- TRAINING GENERATIVE ADVERSARIAL NETWORKS VIA PRIMAL-DUAL SUBGRADIENT METHODS: A LAGRANGIAN PERSPECTIVE ON GAN (ICLR 2018 · similarity 0.91)
- Improving GAN Training with Probability Ratio Clipping and Sample Reweighting (NeurIPS 2020 · similarity 0.91)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.90)
- A Convex Duality Framework for GANs (NeurIPS 2018 · similarity 0.90)
- Improving Generalization and Stability of Generative Adversarial Networks (ICLR 2019 · similarity 0.90)
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.