Research map / Image generation / GANs
Generative Multi-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 |
|---|---|---|---|
| Adding multiple discriminators with automatically regulated softmax (GMAN*) vs single-discriminator GAN | GAN on MNIST · MNIST | GMAM summed score: 0.184 vs 0.005 | Helped |
| Adding multiple discriminators (N=2 to 5) to GMAN-0 (ablation) | GMAN-0 on CIFAR-10 · CIFAR-10 Inception score | 5.878 (N=2) to 5.957 (N=5) | Helped |
| Replacing the modified -log(D) generator objective with the original minimax objective under GMAN | GMAN training · CIFAR-10 Inception score | GMAN-0 5.878 vs mod-GAN 5.738 (N=2) | Helped |
| Generator-controlled lambda (GMAN*) vs fixed lambda in the softmax (ablation) | GMAN with N=5 on MNIST · MNIST GMAM | GMAN* 0.028 vs lambda=0 0.019 and lambda=1 0.008 in pairwise score | Mixed |
| Generator-controlled lambda (GMAN*) vs fixed lambda on CIFAR-10 (ablation) | GMAN with 2 and 5 discriminators · CIFAR-10 Inception score | GMAN* 5.539 vs GMAN-0 5.878 (N=2); 5.955 vs 5.957 (N=5) | Hurt |
| Adding boosting (AdaBoost.OL) to the discriminator (ablation) | GMAN discriminator · MNIST image generation | No clear effect | |
| Increasing number of discriminators from 1 to 5 (GMAN-0) (ablation) | GMAN on MNIST, CelebA, CIFAR-10 · training iterations to steady state | about 2x faster to steady state | Helped |
| Increasing number of discriminators (N=1,2,5) on MNIST (ablation) | GMAN on MNIST · MNIST convergence | reduces iterations to steady-state by 2x | Helped |
Closest papers by meaning
- DO-GAN: A Double Oracle Framework for Generative Adversarial Networks (CVPR 2022 · similarity 0.91)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.91)
- MCL-GAN: Generative Adversarial Networks with Multiple Specialized Discriminators (NeurIPS 2022 · similarity 0.91)
- TRAINING GENERATIVE ADVERSARIAL NETWORKS VIA PRIMAL-DUAL SUBGRADIENT METHODS: A LAGRANGIAN PERSPECTIVE ON GAN (ICLR 2018 · similarity 0.91)
- An Online Learning Approach to Generative Adversarial Networks (ICLR 2018 · similarity 0.90)
- Multi-objective training of Generative Adversarial Networks with multiple discriminators (ICML 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.