Research map / Label-efficient and robust learning / Noisy labels
Meta-Query-Net: Resolving Purity-Informativeness Dilemma in Open-set Active Learning
Dongmin Park, Yooju Shin, Jihwan Bang, Youngjun Lee, Hwanjun Song, Jae-Gil Lee
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 |
|---|---|---|---|
| remove the skyline constraint from the meta-network (ablation) | MQ-Net · CIFAR10, noise 10% | last test accuracy: 93.10 -> 87.25 | Hurt |
| remove the skyline constraint from the meta-network (ablation) | MQ-Net · CIFAR10, noise 60% | last test accuracy: 89.51 -> 81.67 | Hurt |
| replace self-validation set with random labeled validation set (ablation) | MQ-Net · CIFAR10, noise 10% | last test accuracy: 93.10 -> 92.10 | Hurt |
| replace self-validation set with random labeled validation set (ablation) | MQ-Net · CIFAR10, noise 60% | last test accuracy: 89.51 -> 87.65 | Hurt |
| use ReAct instead of CSI as purity score (CONF informativeness) (ablation) | MQ-Net · CIFAR10, noise 40% | last test accuracy: 91.43 -> 89.54 | Hurt |
| use CONF instead of LL as informativeness score (CSI purity) (ablation) | MQ-Net · CIFAR10, noise 40% | last test accuracy: 91.48 -> 91.43 | No clear effect |
| use CONF instead of LL as informativeness score (CSI purity) (ablation) | MQ-Net · CIFAR10, noise 10% | last test accuracy: 93.10 -> 93.28 | No clear effect |
| replace meta-objective with linear combination P(x)+I(x) (ablation) | MQ-Net · CIFAR10, noise 20%, round 10 | test accuracy: 92.10 -> 89.49 | Hurt |
| replace meta-objective with multiplication P(x)*I(x) (ablation) | MQ-Net · CIFAR10, noise 20%, round 10 | test accuracy: 92.10 -> 91.20 | Hurt |
| replace meta-objective with linear combination P(x)+I(x) (ablation) | MQ-Net · CIFAR10, noise 40%, round 10 | test accuracy: 91.48 -> 88.86 | Hurt |
| replace meta-objective with multiplication P(x)*I(x) (ablation) | MQ-Net · CIFAR10, noise 40%, round 10 | test accuracy: 91.48 -> 89.11 | Hurt |
| raise OOD labeling cost c_OOD from 0.5 to 4 | CONF · CIFAR10, noise 40% | last test accuracy: 91.05 -> 80.06 | Hurt |
Closest papers by meaning
- Bidirectional Uncertainty-Based Active Learning for Open-Set Annotation (ECCV 2024 · similarity 0.87)
- Active Sampling for Open-Set Classification without Initial Annotation (AAAI 2019 · similarity 0.87)
- Inconsistency-Based Data-Centric Active Open-Set Annotation (AAAI 2024 · similarity 0.87)
- Rethinking Epistemic and Aleatoric Uncertainty for Active Open-Set Annotation: An Energy-Based Approach (CVPR 2025 · similarity 0.86)
- Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples (ICML 2024 · similarity 0.86)
- Active Classification with Few Queries under Misspecification (NeurIPS 2024 · similarity 0.85)
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.
