Research map / Learning theory and algorithms / Submodular and game algorithms
Do GANs always have Nash equilibria?
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 |
|---|---|---|---|
| proximal training (Algorithm 1) instead of regular GAN training | WGAN-WC and SN-GAN · CIFAR-10 | Inception score: e.g., WGAN-WC DIM=128 2.52 to 4.23; SN-GAN DIM=64 5.12 to 5.72 | Helped |
| optimizing generator against fixed trained discriminator (non-equilibrium check) | SN-GAN regular training · MNIST and CelebA | GAN objective and sample quality | Hurt |
| optimizing proximal objective (lambda=0.1) over generator with fixed discriminator (ablation) | SN-GAN · MNIST and CelebA | GAN objective and sample quality | No clear effect |
Closest papers by meaning
- Finding Mixed Nash Equilibria of Generative Adversarial Networks (ICML 2019 · similarity 0.94)
- On Characterizing GAN Convergence Through Proximal Duality Gap (ICML 2021 · similarity 0.93)
- Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Step (ICLR 2018 · similarity 0.92)
- Generative Adversarial Equilibrium Solvers (ICLR 2024 · similarity 0.90)
- Coulomb GANs: Provably Optimal Nash Equilibria via Potential Fields (ICLR 2018 · similarity 0.90)
- The Numerics of GANs (NeurIPS 2017 · 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.