Research map / Learning theory and algorithms / Sample complexity
Dualing GANs
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 |
|---|---|---|---|
| dual optimization of linear discriminator (instead of saddle-point alternating updates) | GAN with linear discriminator · 5-Gaussians hyperparameter sensitivity | fraction of 100 random hyperparameter settings succeeding in covering 5 modes: 49% vs 32% | Helped |
| dual optimization of linear discriminator | GAN with linear discriminator · MNIST hyperparameter sensitivity | mean Inception score across random hyperparameters: 2.83 vs 1.99 | Helped |
| cost function linearization instead of standard GAN training for nonlinear discriminator | nonlinear discriminator GAN · CIFAR-10 Inception Score (end, own classifier) | Inception score: 5.43 vs 5.61 (standard) | No clear effect |
| score function linearization instead of standard GAN training for nonlinear discriminator | nonlinear discriminator GAN · CIFAR-10 Inception Score (end, own classifier) | Inception score: 5.40 vs 5.61 (standard) | No clear effect |
| cost function linearization for nonlinear discriminator | nonlinear discriminator GAN · CIFAR-10 internal classifier score (end) | internal classifier score: 4.42 vs 3.85 (standard) | Helped |
| score function linearization for nonlinear discriminator | nonlinear discriminator GAN · CIFAR-10 internal classifier score (end) | internal classifier score: 3.52 vs 3.85 (standard) | Hurt |
| dual-based training with linear discriminators, using the dual variables λ | linear discriminator GAN · 5-Gaussians training stability | training objective behavior | Helped |
| increasing minibatch size in dual linear GAN (ablation) | dual linear GAN · 5-Gaussians training stability | training stability | Helped |
| richer (pretrained convnet) features instead of random features in linear discriminator (ablation) | dual linear GAN · MNIST qualitative samples | sample quality | Helped |
Closest papers by meaning
- A Convex Duality Framework for GANs (NeurIPS 2018 · similarity 0.92)
- TRAINING GENERATIVE ADVERSARIAL NETWORKS VIA PRIMAL-DUAL SUBGRADIENT METHODS: A LAGRANGIAN PERSPECTIVE ON GAN (ICLR 2018 · similarity 0.92)
- Dual Discriminator Generative Adversarial Nets (NeurIPS 2017 · similarity 0.92)
- Improving Generalization and Stability of Generative Adversarial Networks (ICLR 2019 · similarity 0.91)
- Hidden Convexity of Wasserstein GANs: Interpretable Generative Models with Closed-Form Solutions (ICLR 2022 · similarity 0.91)
- A Neural Tangent Kernel Perspective of GANs (ICML 2022 · 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.