Research map / Label-efficient and robust learning / Recognition and long-tail
A General Framework for Adversarial Label Learning
Chidubem Arachie, Bert Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 adversarial label learning instead of averaging or GE on weak signals | ALL-3 · Fashion MNIST (SvA) | test accuracy: 0.719 -> 0.924 | Improved |
| use ALL on Cardiotocography against averaging baselines | ALL-3 · Cardiotocography | test accuracy: 0.942 -> 0.803 | Hurt |
| GE instead of ALL on Statlog Satellite | GE-3 · Statlog Satellite | test accuracy: 0.936 -> 0.992 | Mixed |
| add copies of a bad weak signal (ablation) | ALL · Cardiotocography (WS-good plus WS-bad copies) | accuracy | No clear effect |
| add copies of a bad weak signal (ablation) | GE · Cardiotocography (WS-good plus WS-bad copies) | accuracy | Hurt |
| use fixed error bound 0.3 instead of true bounds (ablation) | ALL-3 · Statlog Satellite | test accuracy: 0.936 -> 0.982 | Improved |
| use fixed error bound 0.3 instead of true bounds (ablation) | ALL-3 · Cardiotocography | test accuracy: 0.803 -> 0.657 | Hurt |
| vary fixed error bound between 0 and 1 (ablation) | ALL-3 · Breast Cancer and OBS | test error | No clear effect |
| use reduced BERT embedding features (ablation) | ALL · RTE | label accuracy: 0.834 -> 0.9 | Improved |
| use two-layer neural network learner instead of logistic regression (ablation) | ALL · Word Similarity | label accuracy: 0.933 -> 0.967 | Improved |
| add more crowd annotator labels (ablation) | ALL · Word Similarity | accuracy | Improved |
| add Multi-ALL with precision and error constraints instead of Snorkel | Multi-ALL · Fashion-MNIST (weak signals only) | test error: 0.401 -> 0.335 | Improved |
Closest papers by meaning
- Adversarial Label Learning (AAAI 2019 · similarity 0.97)
- Adversarial Multi Class Learning under Weak Supervision with Performance Guarantees (ICML 2021 · similarity 0.92)
- Adversarial Data Programming: Using GANs to Relax the Bottleneck of Curated Labeled Data (CVPR 2018 · similarity 0.89)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.88)
- Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement (CVPR 2024 · similarity 0.88)
- Semantic-Aware Multi-Label Adversarial Attacks (CVPR 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.
