Research map / Label-efficient and robust learning / Recognition and long-tail
Adversarial Data Programming: Using GANs to Relax the Bottleneck of Curated Labeled Data
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 |
|---|---|---|---|
| Add ADP-generated labeled examples to real training data | ResNet-56 on CIFAR-10 · CIFAR-10 | Test cross-entropy loss at 50k epochs: 4.19 -> 2.71 | Helped |
| Add inter-function dependency modeling to the labeling functions (ablation) | ADP · CIFAR10 | Inception Score: 4.32 -> 8.7 | Helped |
| Use diverse labeling-function types together rather than one type alone (ablation) | ADP · CIFAR10 | Inception Score: 5.52 -> 8.7 | Helped |
| Use heuristic and deep learned feature-based labeling functions together (ablation) | ADP · CIFAR10 | Inception Score: 8.7 -> 7.01 | Hurt |
| Use deep learned feature-based and image processing labeling functions together (ablation) | ADP · CIFAR10 | Inception Score: 8.7 -> 7.52 | Hurt |
| Use image processing and heuristic labeling functions together (ablation) | ADP · CIFAR10 | Inception Score: 8.7 -> 7.27 | Hurt |
| Increase labeling-function count from 3 to 50 (ablation) | ADP · MNIST | Test cross-entropy error: 70.23\% -> 1.33\% | Helped |
| Increase labeling-function count from 3 to 55 (ablation) | ADP · CIFAR10 | Test cross-entropy error: 87.39\% -> 18.45\% | Helped |
| Use 70 instead of 50 labeling functions (ablation) | ADP · MNIST | Test cross-entropy error: 1.33\% -> 1.25\% | Helped |
| Use 55 instead of 50 labeling functions (ablation) | ADP · CIFAR10 | Test cross-entropy error: 18.93\% -> 18.45\% | Helped |
Closest papers by meaning
- A Unified Generative Adversarial Network Training via Self-Labeling and Self-Attention (ICML 2021 · similarity 0.89)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.88)
- Teaching a GAN What Not to Learn (NeurIPS 2020 · similarity 0.88)
- Adversarial Label Learning (AAAI 2019 · similarity 0.88)
- Revisiting Adversarial Training at Scale (CVPR 2024 · similarity 0.87)
- SMaRt: Improving GANs with Score Matching Regularity (ICML 2024 · 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.