Research map / Image generation / GANs
It Takes (Only) Two: Adversarial Generator-Encoder 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 two mappings, an encoder and generator, with an adversarial game between them instead of a separate discriminator | AGE · image generation and reconstruction | model complexity: only two feed-forward mappings | Helped |
| use AGE rather than ALI | CIFAR-10 · CIFAR-10 | Inception score: 5.34 ± 0.05 -> 5.90 ± 0.04 | Helped |
| use AGE rather than ALI | MNIST · MNIST | log likelihood: 721 -> 746 | Helped |
| use AGE rather than ALI | SVHN unsupervised feature learning · SVHN | error rate: 19.14% -> 23.7% | Hurt |
| add data-space reconstruction loss to the generator objective (ablation) | AGE · image generation | generator output quality: degradation | Hurt |
| remove reconstruction terms (ablation) | AGE · CIFAR-10 | sample diversity: diverse samples | No clear effect |
| condition the encoder and generator on the grayscale image | conditional AGE · Stanford Cars colorization | colorization diversity | Helped |
| add three noise maps to pix2pix generator input | pix2pix · Stanford Cars colorization | colorization diversity: diversity of colorizations was very low | No clear effect |
Closest papers by meaning
- Adversarially Learned Inference (ICLR 2017 · similarity 0.90)
- Generalized Adversarially Learned Inference (AAAI 2021 · similarity 0.89)
- DO-GAN: A Double Oracle Framework for Generative Adversarial Networks (CVPR 2022 · similarity 0.88)
- Structured Generative Adversarial Networks (NeurIPS 2017 · similarity 0.88)
- Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks (ICML 2017 · similarity 0.87)
- SGAN: An Alternative Training of Generative Adversarial Networks (CVPR 2018 · similarity 0.87)
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.