Research map / Label-efficient and robust learning / Recognition and long-tail
Contrastive Coding for Active Learning Under Class Distribution Mismatch
Pan Du, Suyun Zhao, Hui Chen, Shuwen Chai, Hong Chen, Cuiping Li
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 semantic score Ssem to the query (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy | Improved |
| add distinctive score Sdis to the query (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy | Improved |
| learn semantic features with contrastive learning instead of VAE (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy: +3.74% at 5 AL cycles | Improved |
| learn distinctive features with VAE instead of contrastive learning (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy | No clear effect |
| raise threshold t from 0.7 to 0.9 (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy | Improved |
| use k=70 instead of k=130 (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy | No clear effect |
| use k=10 instead of larger k (ablation) | CCAL · CIFAR10 (60% mismatch) | accuracy | Hurt |
| use CCAL instead of random sampling | random sampling · CIFAR10 (80% mismatch, 5 AL cycles) | accuracy: +2.44% | Improved |
| use CCAL instead of random sampling | random sampling · CIFAR100 (80% mismatch, 5 AL cycles) | accuracy: +7.65% | Improved |
| increase class mismatch ratio | existing AL methods (VAAL, Coreset) · CIFAR10/CIFAR100 | accuracy | Hurt |
| use CCAL (1 AL cycle) instead of SSL methods DS3L and UASD | DS3L, UASD · CIFAR10 | accuracy | Improved |
Closest papers by meaning
- Formal Models of Active Learning from Contrastive Examples (NeurIPS 2025 · similarity 0.85)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.85)
- CoLAL: Co-learning Active Learning for Text Classification (AAAI 2024 · similarity 0.84)
- Universal Rates for Active Learning (NeurIPS 2024 · similarity 0.84)
- Unlocking the Power of Open Set: A New Perspective for Open-Set Noisy Label Learning (AAAI 2024 · similarity 0.84)
- Unsupervised Learning for Class Distribution Mismatch (ICML 2025 · similarity 0.84)
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.
