Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
MSplit LBI: Realizing Feature Selection and Dense Estimation Simultaneously in Few-shot and Zero-shot 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 the dense estimator β, which captures weak signals in addition to strong signals (ablation) | MSplit LBI · Simulation (σ=0.2, 0.4, 0.6, 0.8) | relative estimation error: 0.1463 ± 0.0241 -> 0.1238 ± 0.0112; 0.1714 ± 0.0255 -> 0.1312 ± 0.0117; 0.2016 ± 0.0206 -> 0.1461 ± 0.0120; 0.2421 ± 0.0 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · AwA | accuracy (%): 84.58 -> 85.34 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · CUB | accuracy (%): 56.62 -> 57.52 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · ImageNet Top-1 | accuracy (%): 7.98 -> 8.35 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · ImageNet Top-5 | accuracy (%): 17.83 -> 18.76 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · Omniglot 5-way 1-shot | accuracy (%): 94.7 -> 94.8 | No clear effect |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · Omniglot 5-way 5-shot | accuracy (%): 98.9 -> 99.2 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · Omniglot 20-way 1-shot | accuracy (%): 83.7 -> 83.7 | No clear effect |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · Omniglot 20-way 5-shot | accuracy (%): 97.3 -> 97.6 | Helped |
| use the dense estimator β instead of sparse estimator β̃ (ablation) | MSplit LBI · SUN 5-way 1-shot | accuracy (%): 59.32 -> 61.47 | Helped |
| fine-tune the pretrained feature model (ablation) | C-Siam · Omniglot | accuracy (%): 96.7 -> 97.3; 98.4 -> 98.4; 88.0 -> 88.1; 96.5 -> 97.0 | Mixed |
| fine-tune the pretrained feature model (ablation) | M-Net · Omniglot | accuracy (%): 98.1 -> 97.9; 98.9 -> 98.7; 93.8 -> 93.5; 98.5 -> 98.7 | Mixed |
Closest papers by meaning
- LP++: A Surprisingly Strong Linear Probe for Few-Shot CLIP (CVPR 2024 · similarity 0.83)
- Subspace Regularizers for Few-Shot Class Incremental Learning (ICLR 2022 · similarity 0.83)
- Boosting Few-Shot Learning via Attentive Feature Regularization (AAAI 2024 · similarity 0.82)
- Learning Sparse Neural Networks through L_0 Regularization (ICLR 2018 · similarity 0.82)
- A Universal Analysis of Large-Scale Regularized Least Squares Solutions (NeurIPS 2017 · similarity 0.82)
- Meta-Learning With Differentiable Convex Optimization (CVPR 2019 · similarity 0.82)
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.