Research map / Label-efficient and robust learning / Knowledge distillation
Few Shot Network Compression via Cross Distillation
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 cross distillation to layer-wise pruning distillation (ablation) | VGG-16 · CIFAR-10, structured pruning, VGG-50%, 1 sample/class | top-1 accuracy (%): 65.57_{\pm 1.61} -> 69.25_{\pm 1.39} | Helped |
| Use soft cross distillation instead of no cross distillation (ablation) | VGG-16 · ILSVRC-12, unstructured pruning, VGG-90%, 1 sample/class | top-5 accuracy (%): 85.21_{\pm 0.41} -> 88.14_{\pm 0.74} | Helped |
| Add cross distillation to quantization (ablation) | VGG-16 · CIFAR-10, W2A3 quantization, 1 sample/class | accuracy (%): 53.24_{\pm 2.31} -> 60.36_{\pm 2.86} | Helped |
| Add Mixup to cross distillation (ablation) | ResNet-56 · CIFAR-10, structured pruning, 1 sample/class | accuracy (%): 69.25_{\pm 1.39} -> 79.97_{\pm 1.71} | Helped |
| Use cross connections only at deeper VGG layers rather than early layers (ablation) | Ours-S · CIFAR-10, VGG-16 structured pruning, VGG-50%, 10 samples/class | accuracy (%): C1.2: 83.98_{\pm 0.63}; C5.3: 84.97_{\pm 71} | Mixed |
Closest papers by meaning
- Progressive Network Grafting for Few-Shot Knowledge Distillation (AAAI 2021 · similarity 0.91)
- Few Sample Knowledge Distillation for Efficient Network Compression (CVPR 2020 · similarity 0.91)
- Model compression via distillation and quantization (ICLR 2018 · similarity 0.87)
- Cross-Level Distillation and Feature Denoising for Cross-Domain Few-Shot Classification (ICLR 2023 · similarity 0.87)
- EPSD: Early Pruning with Self-Distillation for Efficient Model Compression (AAAI 2024 · similarity 0.86)
- Layer Compression of Deep Networks with Straight Flows (AAAI 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.