Research map / Label-efficient and robust learning / Recognition and long-tail
Active Learning with Partial Feedback
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 |
|---|---|---|---|
| Use ALPF with ERC rather than random-example baseline annotation | ResNet-18 · TinyImageNet, 30% annotation budget | top-1 accuracy: 0.310 -> 0.391 | Helped |
| Use ALPF with ERC rather than AQ with EDC | ResNet-18 · TinyImageNet, 30% annotation budget | top-1 accuracy: 0.353 -> 0.391 | Helped |
| Use ALPF with ERC rather than random-example baseline annotation | ResNet-18 · CIFAR100, 30% annotation budget | top-1 accuracy: 0.412 -> 0.464 | Helped |
| Use ALPF with EDC rather than random-example baseline annotation | ResNet-18 · CIFAR10, 30% annotation budget | top-1 accuracy: 0.757 -> 0.797 | Helped |
| Use ERC to select questions and examples instead of baseline annotation | ResNet-18 · TinyImageNet | labeling cost: 827k -> 491k | Helped |
| Use ERC to select questions and examples instead of baseline annotation | ResNet-18 · CIFAR100 | labeling cost: 337k -> 187k | Helped |
| Use EDC to select questions and examples instead of baseline annotation | ResNet-18 · CIFAR10 | labeling cost: 170k -> 74k | Helped |
| Use conventional uncertainty-based active learning rather than random sampling | ResNet-18 · TinyImageNet | top-1 accuracy: 0.310; 0.303; 0.313 | No clear effect |
| Add partial labels to training rather than exclude them (ablation) | multiclass classifier · TinyImageNet | holdout accuracy | Helped |
| Make partial labels coarser (ablation) | multiclass classifier trained with partial labels · TinyImageNet | holdout accuracy | Hurt |
| Use no i.i.d. warm-start data instead of 5% warm-starting (ablation) | ALPF - EIG and ALPF - EDC · evaluated datasets | top-1 accuracy | Hurt |
| Increase warm-starting from 5% to 10% (ablation) | ALPF learners · evaluated datasets | top-1 accuracy | No clear effect |
Closest papers by meaning
- Re-Active Learning: Active Learning with Relabeling (AAAI 2016 · similarity 0.88)
- Active Learning for Semantic Segmentation with Multi-class Label Query (NeurIPS 2023 · similarity 0.88)
- Active Learning of Multi-Class Classification Models from Ordered Class Sets (AAAI 2019 · similarity 0.87)
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 2024 · similarity 0.87)
- CoLAL: Co-learning Active Learning for Text Classification (AAAI 2024 · similarity 0.85)
- Instance-wise Supervision-level Optimization in Active Learning (CVPR 2025 · 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.