Research map / Image generation / GANs
A Relaxed Inertial Forward-Backward-Forward Algorithm for Solving Monotone Inclusions with Application to GANs
Radu I. Bot, Michael Sedlmayer, Phan Tu Vuong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 inertial term (alpha = 0.05) to FBF Adam (ablation) | FBF Adam · CIFAR10 (WGAN, DCGAN) | FID: 45.85 -> 45.25 | Improved |
| add inertial term (alpha = 0.05) to FBF Adam (ablation) | FBF Adam · CIFAR10 (WGAN, DCGAN) | Inception score: 4.54 -> 4.59 | Improved |
| use FBF Adam instead of Extra Adam | WGAN training · CIFAR10 (WGAN, DCGAN) | FID: 56.67 -> 45.85 | Improved |
| use FBF Adam instead of PDSP Adam | WGAN training · CIFAR10 (WGAN, DCGAN) | FID: 53.97 -> 45.85 | Improved |
| increase inertial parameter alpha toward its upper limit (ablation) | RIFBF on bilinear saddle point · Bilinear saddle point (m=n=500, mu=0.5) | iterations to residual <= 1e-5 | Improved |
| increase relaxation parameter rho, including over-relaxation rho > 1 (ablation) | RIFBF on bilinear saddle point · Bilinear saddle point (m=n=500, mu=0.5) | iterations to residual <= 1e-5 | Improved |
| use larger stepsize factor mu (0.9 or 0.5 rather than 0.1) (ablation) | RIFBF on bilinear saddle point · Bilinear saddle point (m=n=500) | iterations to residual <= 1e-5 | Improved |
Closest papers by meaning
- First-order Convergence Theory for Weakly-Convex-Weakly-Concave Min-max Problems (JMLR 2021 · similarity 0.82)
- A Fast Optimistic Method for Monotone Variational Inequalities (ICML 2023 · similarity 0.81)
- Optimistic Dual Extrapolation for Coherent Non-monotone Variational Inequalities (NeurIPS 2020 · similarity 0.81)
- Gradient descent GAN optimization is locally stable (NeurIPS 2017 · similarity 0.81)
- Smoothness and Stability in GANs (ICLR 2020 · similarity 0.80)
- Optimistic mirror descent in saddle-point problems: Going the extra (gradient) mile (ICLR 2019 · similarity 0.80)
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.
