Research map / Image generation / GANs
Triangle 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 Δ-GAN instead of Triple GAN | CIFAR10 semi-supervised classification · CIFAR10 (n=4000 labels) | classification error: 16.99 ± 0.36 -> 16.80 ± 0.42 | Helped |
| Use Δ-GAN instead of Triple GAN | CIFAR10 semi-supervised classification · CIFAR10 (n=100 labels) | classification accuracy: 63.79 { ± 0.85} -> 83.20{ ± 1.88} | Helped |
| Use Δ-GAN instead of Triple GAN | CIFAR10 semi-supervised classification · CIFAR10 (n=1000 labels) | classification accuracy: 84.93 { ± 1.63} -> 88.98{ ± 1.50} | Helped |
| Use Δ-GAN instead of Triple GAN | CIFAR10 semi-supervised classification · CIFAR10 (all labels) | classification accuracy: 86.70 { ± 1.52} -> 93.34{ ± 1.46} | Helped |
| Add paired domain supervision using Δ-GAN rather than DiscoGAN | edges2shoes translation · edges2shoes | MSE: 140.1 -> 125.3 | Helped |
| Increase the paired-data fraction from 10% to 20% | Δ-GAN · edges2shoes | MSE: 125.3 -> 113.0 | Helped |
| Increase the paired-data fraction from 20% to 100% | Δ-GAN · edges2shoes | MSE: 113.0 -> 66.4 | Helped |
| Use Δ-GAN instead of Triple GAN | CelebA attribute generation · CelebA (1% labeled) | P@10: 40.97 -> 53.21 | Helped |
| Use Δ-GAN instead of Triple GAN | COCO attribute generation · COCO (50% labeled) | nDCG@10: 37.76 -> 40.39 | Helped |
| Replace Δ-GAN with Triple GAN-s | toy joint-distribution matching · Toy mixture data | joint distribution matching | Hurt |
Closest papers by meaning
- Triple Generative Adversarial Nets (NeurIPS 2017 · similarity 0.90)
- JointGAN: Multi-Domain Joint Distribution Learning with Generative Adversarial Nets (ICML 2018 · similarity 0.89)
- Conditional GANs with Auxiliary Discriminative Classifier (ICML 2022 · similarity 0.88)
- Twin Auxilary Classifiers GAN (NeurIPS 2019 · similarity 0.88)
- Good Semi-supervised Learning That Requires a Bad GAN (NeurIPS 2017 · similarity 0.88)
- Dual Discriminator Generative Adversarial Nets (NeurIPS 2017 · 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.