Research map / Architectures and efficiency / Pruning and efficient tuning
Efficient K-Shot Learning With Regularized Deep Networks
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 |
|---|---|---|---|
| Add activation-based parameter clustering (ablation) | Fine-tuning with ResNet-18 · Office Amazon-to-Webcam, 1-shot | accuracy (%): 70.07 -> 79.94 | Helped |
| Add margin loss to GNA (ablation) | GNA with ResNet-18 · Office Amazon-to-Webcam, 1-shot | accuracy (%): 79.94 -> 82.16 | Helped |
| Use RL rather than greedy grouping search (ablation) | GNA + margin loss · Office Amazon-to-Webcam, 1-shot | accuracy (%): 83.16 -> 85.04 | Helped |
| Add margin loss to standard fine-tuning (ablation) | ResNet-18 fine-tuning · Office Amazon-to-Webcam, 1-shot | accuracy (%): 70.07 -> 70.34 | Helped |
| Add activation-based parameter clustering (ablation) | Fine-tuning with ResNet-18 · CIFAR-100 to CIFAR-10, 1-shot | accuracy (%): 29.58 -> 32.70 | Helped |
| Add margin loss to GNA (ablation) | GNA with ResNet-18 · CIFAR-100 to CIFAR-10, 1-shot | accuracy (%): 32.70 -> 34.43 | Helped |
| Use RL rather than greedy grouping search (ablation) | GNA + margin loss · CIFAR-100 to CIFAR-10, 1-shot | accuracy (%): 33.50 -> 35.95 | Helped |
| Add activation-based parameter clustering (ablation) | Fine-tuning with ResNet-18 · CIFAR-100 to CIFAR-10, 10-shot | accuracy (%): 56.00 -> 57.96 | Helped |
| Add triplet loss to standard fine-tuning (ablation) | ResNet-18 fine-tuning · CIFAR-100 to CIFAR-10, 10-shot | accuracy (%): 56.00 -> 58.17 | Helped |
| Use RL rather than greedy grouping search (ablation) | GNA + triplet loss · CIFAR-100 to CIFAR-10, 10-shot | accuracy (%): 58.56 -> 60.30 | Helped |
| Add parameter clustering (ablation) | Fine-tuning · k-shot classification, 25-shot | accuracy (%): 80.41 -> 84.48 | Helped |
| Add parameter clustering (ablation) | Fine-tuning · k-shot classification, 5-shot | accuracy (%): 68.84 -> 68.93 | Helped |
Closest papers by meaning
- Distance-Based Regularisation of Deep Networks for Fine-Tuning (ICLR 2021 · similarity 0.87)
- AdaFilter: Adaptive Filter Fine-Tuning for Deep Transfer Learning (AAAI 2020 · similarity 0.86)
- Growing a Brain: Fine-Tuning by Increasing Model Capacity (CVPR 2017 · similarity 0.86)
- Intraclass clustering: an implicit learning ability that regularizes DNNs (ICLR 2021 · similarity 0.86)
- Neural Fine-Tuning Search for Few-Shot Learning (ICLR 2024 · similarity 0.85)
- Quick-Tune: Quickly Learning Which Pretrained Model to Finetune and How (ICLR 2024 · 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.