Research map / Image generation / GANs
An Online Learning Approach to 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 |
|---|---|---|---|
| Use past generator and discriminator states in Chekhov GAN | DCGAN on Stacked MNIST · Stacked MNIST | Generated Classes: 629 ± 121.08 -> 743 ± 64.31 | Helped |
| Use past generator and discriminator states in Chekhov GAN | DCGAN on Stacked MNIST · Stacked MNIST | Reverse KL: 1.96 ± 0.64 -> 1.40 ± 0.21 | Helped |
| Increase past states from 5 to 10 (ablation) | Chekhov GAN on Stacked MNIST · Stacked MNIST | Generated Classes: 743 ± 64.31 -> 795 ± 37 | Helped |
| Increase past states from 5 to 10 (ablation) | Chekhov GAN on Stacked MNIST · Stacked MNIST | Reverse KL: 1.40 ± 0.21 -> 1.24 ± 0.17 | Helped |
| Use 25 past states | Chekhov GAN on CIFAR10 · CIFAR10 train set | MSE: 61.13 ± 3.99 -> 48.42 ± 2.99 | Helped |
| Use 25 past states | Chekhov GAN on CIFAR10 · CIFAR10 test set | MSE: 59.5 ± 3.65 -> 46.82 ± 2.96 | Helped |
| Use 25 past states | Chekhov GAN on CIFAR10 · CIFAR10 train set | Best Rank (%): 0 % -> 81.33 % | Helped |
| Use 25 past states | Chekhov GAN on CIFAR10 · CIFAR10 test set | Best Rank (%): 0 % -> 82.43 % | Helped |
| Use 5 past states | Chekhov GAN on CelebA · CelebA, σ=0.25 | Estimated missing modes: 3004 ± 4154 -> 1407 ± 1848 | Helped |
| Use 5 past states | Chekhov GAN on CelebA · CelebA, σ=0.5 | Estimated missing modes: 2568.25 ± 4148 -> 1007 ± 1805 | Helped |
| Reduce latent dimension to match the data dimensionality (ablation) | GAN and Chekhov GAN on the Gaussian mixture toy dataset · Mixture of Gaussians | Mode coverage | Helped |
| Use Chekhov GAN instead of vanilla GAN | GAN on the imbalanced 5-Gaussian mixture · Mixture of Gaussians | Mode coverage | Helped |
Closest papers by meaning
- The Numerics of GANs (NeurIPS 2017 · similarity 0.93)
- DO-GAN: A Double Oracle Framework for Generative Adversarial Networks (CVPR 2022 · similarity 0.92)
- Finding Mixed Nash Equilibria of Generative Adversarial Networks (ICML 2019 · similarity 0.91)
- Training Generative Adversarial Networks in One Stage (CVPR 2021 · similarity 0.91)
- Towards Principled Methods for Training Generative Adversarial Networks (ICLR 2017 · similarity 0.91)
- Generative Multi-Adversarial Networks (ICLR 2017 · 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.