Research map / Label-efficient and robust learning / Semi-supervised learning
Enhancing TripleGAN for Semi-Supervised Conditional Instance Synthesis and Classification
Check 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 |
|---|---|---|---|
| Add feature-semantics matching to the generator (ablation) | EnhancedTGAN · CIFAR-10, 4000 labels | classification accuracy on synthesized data | Helped |
| Add semantics matching to the generator (ablation) | EnhancedTGAN · CIFAR-10, 4000 labels | classification accuracy on synthesized data | Helped |
| Add synthesized instances by retaining the generator and discriminator (ablation) | EnhancedTGAN classifiers · CIFAR-10, 4000 labels | test error rate (%): 11.92±0.19 -> 9.42±0.22 | Helped |
| Add consistency regularization between classifiers (ablation) | EnhancedTGAN · CIFAR-10, 4000 labels | test error rate (%): 11.47±0.13 -> 9.42±0.22 | Helped |
| Add smoothness regularization (ablation) | EnhancedTGAN · CIFAR-10, 4000 labels | test error rate (%): 12.03±0.29 -> 9.42±0.22 | Helped |
| Replace TripleGAN∗ with EnhancedTGAN | semi-supervised classification · MNIST, 100 labels | test error rate (%): 0.81±0.08 -> 0.42±0.03 | Helped |
| Replace TripleGAN∗ with EnhancedTGAN | semi-supervised classification · SVHN, 1000 labels | test error rate (%): 4.53±0.22 -> 2.97±0.09 | Helped |
| Replace TripleGAN∗ with EnhancedTGAN | semi-supervised classification · CIFAR-10, 4000 labels | test error rate (%): 14.65±0.38 -> 9.42±0.22 | Helped |
| Replace TripleGAN∗ with EnhancedTGAN | semi-supervised classification · FaceScrub-100, 2000 labels | test error rate (%): 18.23±0.56 -> 16.08±0.24 | Helped |
Closest papers by meaning
- Regularizing Discriminative Capability of CGANs for Semi-Supervised Generative Learning (CVPR 2020 · similarity 0.91)
- Triple Generative Adversarial Nets (NeurIPS 2017 · similarity 0.91)
- MarginGAN: Adversarial Training in Semi-Supervised Learning (NeurIPS 2019 · similarity 0.89)
- Good Semi-supervised Learning That Requires a Bad GAN (NeurIPS 2017 · similarity 0.89)
- A Unified View of cGANs with and without Classifiers (NeurIPS 2021 · similarity 0.88)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 2024 · similarity 0.88)
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.