Research map / Label-efficient and robust learning / Recognition and long-tail
Active Learning by Feature Mixing
Check 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 ALFA-Mix feature mixing and candidate clustering | MViT on HMDB · HMDB, first two AL rounds | Top-1 test accuracy: more than $5\%$ | Helped |
| Use ALFA-Mix feature mixing and candidate clustering | MViT on HMDB · HMDB, later AL rounds | Top-1 test accuracy: $3\%$ | Helped |
| Use ALFA-Mix instead of the strongest baselines | MViT on HMDB · HMDB, first three AL rounds | Top-1 test accuracy: more than $2\%$ | Helped |
| Use k-means clustering for candidate selection (ablation) | ALFA-Mix candidate set · Diversification experiments | Test performance | Helped |
| Replace k-means with k-means++ (ablation) | ALFA-Mix candidate set · Diversification experiments | Test performance | Hurt |
| Learn the interpolation parameter instead of fixing or skipping its optimization (ablation) | ALFA-Mix · Candidate selection | Candidate-set size | Helped |
| Use class anchors from labelled data (ablation) | ALFA-Mix · Anchor ablation experiments | Test performance | Helped |
| Interpolate with labelled anchors rather than adding adversarial noise (ablation) | ALFA-Mix · Latent-space perturbation comparison | Test performance | Helped |
| Use closed-form interpolation optimization instead of iterative gradient optimization (ablation) | ALFA-Mix · Interpolation optimization | Acquisition time: 5x slower | Helped |
| Use closed-form interpolation optimization instead of iterative gradient optimization (ablation) | ALFA-Mix · Interpolation optimization | Test performance | No clear effect |
| Use ALFA-Mix rather than BADGE | ALFA-Mix · Small-budget experiments | Test performance: $46\%$ | Helped |
| Use ALFA-Mix rather than other active-learning approaches | ALFA-Mix · Low-data regime, small budgets | Test performance: at least $60\%$ | Helped |
Closest papers by meaning
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 2024 · similarity 0.89)
- Active Learning for Convolutional Neural Networks: A Core-Set Approach (ICLR 2018 · similarity 0.87)
- Un-mix: Rethinking Image Mixtures for Unsupervised Visual Representation Learning (AAAI 2022 · similarity 0.87)
- SIMILAR: Submodular Information Measures Based Active Learning In Realistic Scenarios (NeurIPS 2021 · similarity 0.87)
- Instance-wise Supervision-level Optimization in Active Learning (CVPR 2025 · similarity 0.86)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.86)
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.