Research map / Architectures and efficiency / RNNs and normalization
Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-Learning
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 discrete two-step pruning instead of continuous pruning (ablation) | MetaDOCK on CIFAR-fs · CIFAR-fs, 5-way 5-shot, budget 25% | test accuracy (%): 67.78 ± 0.19 -> 69.30 ± 0.19 | Helped |
| Use discrete two-step pruning instead of continuous pruning (ablation) | MetaDOCK on CIFAR-fs · CIFAR-fs, 5-way 5-shot, budget 12.5% | test accuracy (%): 64.87 ± 0.20 -> 70.15 ± 0.19 | Helped |
| Adapt kernel masks separately for each task rather than pruning only globally (ablation) | MetaDOCK on CIFAR-fs · CIFAR-fs, 5-way 5-shot, budget about 13% | test accuracy (%): 69.98 ± 0.19 -> 70.15 ± 0.19 | Helped |
| Adapt kernel masks separately for each task rather than pruning only globally (ablation) | MetaDOCK on CIFAR-fs · CIFAR-fs, 5-way 5-shot, budget about 7% | test accuracy (%): 69.92 ± 0.19 -> 70.21 ± 0.19 | Helped |
| Add dynamic task-specific kernel pruning | iMAML with 4-Conv-128 · CIFAR-fs, 5-way 5-shot | test accuracy (%): 68.08 ± 0.19 -> 70.21 ± 0.19 | Helped |
| Add dynamic task-specific kernel pruning at a moderate budget | iMAML with 4-Conv-128 · mini-ImageNet, 5-way 5-shot | test accuracy (%): 63.39 ± 0.18 -> 64.34 ± 0.18 | Helped |
| Prune below 10% budget (ablation) | MetaDOCK on mini-ImageNet · mini-ImageNet, few-shot classification | test accuracy | Mixed |
| Reduce the pruning budget (ablation) | MetaDOCK task-specific models · CIFAR-fs and mini-ImageNet | Measure of Meta-Overfitting (MO) | Helped |
| Prune a 4-Conv-128 model to 25% budget rather than use an unpruned 4-Conv-64 model (ablation) | MetaDOCK on CIFAR-fs · CIFAR-fs, 5-way 1-shot | test accuracy (%): 67.25% -> 70.15% | Helped |
Closest papers by meaning
- Learning Large-scale Neural Fields via Context Pruned Meta-Learning (NeurIPS 2023 · similarity 0.90)
- Learning To Learn by Jointly Optimizing Neural Architecture and Weights (CVPR 2022 · similarity 0.89)
- Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation (ICML 2021 · similarity 0.89)
- Meta-Learning with Neural Tangent Kernels (ICLR 2021 · similarity 0.89)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.89)
- MetaOptimize: A Framework for Optimizing Step Sizes and Other Meta-parameters (ICML 2025 · similarity 0.87)
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.