Research map / Learning theory and algorithms / Prediction and decision losses
Discover-Then-Rank Unlabeled Support Vectors in the Dual Space for Multi-Class Active Learning
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 D-TRUST (discover-then-rank dual-space sampling) instead of baseline AL samplers | SVM active learner · Dermatology1, USPS, Penstroke (Table 4 small settings) | test accuracy: e.g., 77.0 vs 76.0 (Derm1 @100); 95.3 vs 94.7 (USPS @500) | Helped |
| Add uncertainty assistance (UNpair and UNall terms) to the Fdisc discovery function | D-TRUST discovery phase · Real datasets | test accuracy | No clear effect |
| Increase large C (margin constraint) value (ablation) | SVM in D-TRUST · Synthetic data | model accuracy: 0.99 (C=1,10) to 0.98 (C=100) | Hurt |
| Use a transition parameter eta that is too small or too large (switching discovery to ranking too early or too late) (ablation) | D-TRUST transition criterion · Dermatology 2, Yeast, Dermatology 1, Auto-Drive | test accuracy | Mixed |
| Vary lambda1 and gamma1 balancing weights by up to 5x (ablation) | D-TRUST sampling score · Dermatology2, Auto-drive, USPS | test accuracy | No clear effect |
| Use a 10-sample batch-mode variant of D-TRUST | D-TRUST · Yeast, USPS, Dermatology1, Penstroke | test accuracy | No clear effect |
| Replace SVM with LR or KNN classifier for QUIRE and GD baselines (ablation) | QUIRE and Graph Density sampling · Dermatology, Yeast, USPS, Penstroke, Auto-drive | test accuracy | Hurt |
Closest papers by meaning
- Integrating Bayesian and Discriminative Sparse Kernel Machines for Multi-class Active Learning (NeurIPS 2019 · similarity 0.88)
- A Lagrangian Duality Approach to Active Learning (NeurIPS 2022 · similarity 0.88)
- Transductive Active Learning: Theory and Applications (NeurIPS 2024 · similarity 0.86)
- Improved Active Learning via Dependent Leverage Score Sampling (ICLR 2024 · similarity 0.86)
- Querying Easily Flip-flopped Samples for Deep Active Learning (ICLR 2024 · similarity 0.85)
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 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.