Research map / Label-efficient and robust learning / Domain adaptation and generalization
argmax centroid
Chengyue Gong, Mao Ye, qiang liu
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 argmax centroid heads on top of SIB | SIB · Mini-ImageNet 1-shot | accuracy: 70.0 -> 71.2 | Improved |
| add argmax centroid heads on top of SIB | SIB · Mini-ImageNet 5-shot | accuracy: 80.0 -> 81.1 | Improved |
| add argmax centroid heads on top of IFSL | IFSL · Mini-ImageNet 1-shot | accuracy: 73.2±0.3 -> 73.6±0.4 | Improved |
| add argmax centroid heads on top of SIB | SIB · TieredImageNet 1-shot | accuracy: 81.2±0.6 -> 81.8±0.6 | Improved |
| add argmax centroid heads on top of IFSL | IFSL · TieredImageNet 5-shot | accuracy: 88.3±0.5 -> 88.5±0.5 | Improved |
| select best of n heads instead of averaging ensemble outputs | Ours · Mini-ImageNet 1-shot ResNet-18 | accuracy: 56.7±.3 -> 57.4±.2 | Improved |
| select best of n heads vs Bayesian MAML | Ours · Mini-ImageNet 1-shot WRN-28-10 | accuracy: 57.7±.4 -> 58.1±.4 | Improved |
| centroid heads for personalized dialogue vs PAML | Ours · personalized dialogue | PPL: 46.35 -> 43.94 | Improved |
| centroid heads for personalized dialogue vs PAML | Ours · personalized dialogue | BLEU: 0.77 -> 0.91 | Improved |
| domain-specific adaptor instead of one shared adaptor | domain adaptation baseline · DomainNet subset | accuracy: 39.5 -> 41.8 | Improved |
| argmax centroid adaptors instead of one adaptor per domain | Ours · DomainNet subset | accuracy: 41.8 -> 44.3 | Improved |
| random domain-to-model assignment with ensembling | Ensemble · DomainNet subset | accuracy: 41.8 -> 41.3 | Hurt |
Closest papers by meaning
- Centroid Estimation With Guaranteed Efficiency: A General Framework for Weakly Supervised Learning (TPAMI 2020 · similarity 0.85)
- Alleviating the Sample Selection Bias in Few-shot Learning by Removing Projection to the Centroid (NeurIPS 2022 · similarity 0.84)
- An Efficient Multi-Estimation-Based Parameter Centroid Decision via Linear Regression Approach (TPAMI 2026 · similarity 0.84)
- Improving Task-Specific Generalization in Few-Shot Learning via Adaptive Vicinal Risk Minimization (NeurIPS 2022 · similarity 0.83)
- MOKD: Cross-domain Finetuning for Few-shot Classification via Maximizing Optimized Kernel Dependence (ICML 2024 · similarity 0.83)
- Multi-Centroid Task Descriptor for Dynamic Class Incremental Inference (CVPR 2023 · similarity 0.83)
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.
