Research map / Label-efficient and robust learning / Recognition and long-tail
Active Decision Boundary Annotation With Deep Generative Models
Miriam Huijser, Jan C. van Gemert
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 |
|---|---|---|---|
| annotate decision boundary points instead of only sample labels | uncertainty sampling · SVHN 0 vs. 8 | AULC: 118.7 -> 124.3 | Improved |
| annotate decision boundary points instead of only sample labels | uncertainty-dense sampling · Shoe-Bag | AULC: 112.0 -> 135.2 | Improved |
| annotate decision boundary points instead of only sample labels | 5 cluster centroid · SVHN 0 vs. 8 | AULC: 98.0 -> 106.3 | Improved |
| annotate decision boundary points instead of only sample labels | random sampling · MNIST 0 vs. 8 | AULC: 142.2 -> 145.1 | Improved |
| use uncertainty sampling rather than other query strategies (ablation) | active learning · MNIST 0 vs. 8 | AULC: 135.6 -> 144.0 | Improved |
| add Gaussian noise (sigma 5 images) to boundary annotations (ablation) | boundary annotation · SVHN 0 vs. 8 | AULC: 124.0 -> 113.6 | Hurt |
| add Gaussian noise (sigma 1 to 4 images) to boundary annotations (ablation) | boundary annotation · MNIST 0 vs. 8 | AULC: 146.0 -> 144.2 | Hurt |
| use human oracle instead of SVM oracle for boundary annotation (ablation) | boundary annotation · MNIST 0 vs. 8 (10 queries) | AULC: 9.1 -> 8.8 | Hurt |
| annotate decision boundary with human oracle instead of sample labels | active learning · SVHN 0 vs. 8 (10 queries) | AULC: 5.7 -> 5.8 | No clear effect |
| annotate decision boundary instead of samples, averaged over all class pairs | active learning · SVHN (all class pairs) | AULC: 127.8 -> 130.9 | Improved |
| use a lower quality generative model (Shoe-Bag embedding) (ablation) | human annotation consistency · Shoe-Bag | lines without change: 5 of 10 lines without change | Hurt |
Closest papers by meaning
- Contextual Active Model Selection (NeurIPS 2024 · similarity 0.84)
- Boundary Matters: A Bi-Level Active Finetuning Method (NeurIPS 2024 · similarity 0.83)
- Inconsistency-Based Data-Centric Active Open-Set Annotation (AAAI 2024 · similarity 0.83)
- Instance-wise Supervision-level Optimization in Active Learning (CVPR 2025 · similarity 0.83)
- Probing the Decision Boundaries of In-context Learning in Large Language Models (NeurIPS 2024 · similarity 0.82)
- Rethinking Epistemic and Aleatoric Uncertainty for Active Open-Set Annotation: An Energy-Based Approach (CVPR 2025 · similarity 0.82)
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.
