Research map / Architectures and efficiency / Pruning and efficient tuning
Neural Epitome Search for Architecture-Agnostic Network Compression
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 |
|---|---|---|---|
| Replace fixed-stride WSNet compression with learned epitome transformations | 8-layer CNN · ESC-50 | accuracy: 6.5% | Helped |
| Use NES rather than the full MobileNetV2 baseline | MobileNetV2 · ImageNet | Top-1 accuracy: 1.47% | Helped |
| Use NES rather than MobileNetV2-0.35 | MobileNetV2-0.35 · ImageNet | Top-1 accuracy: 3.78% | Helped |
| Use NES compression with MobileNetV2-0.7 rather than AMC | MobileNetV2-0.7 · ImageNet | Top-1 accuracy: 2.47% | Helped |
| Apply NES to EfficientNet-B0 | EfficientNet-B0 · ImageNet | Top-1 accuracy: 1.2% | Helped |
| Use NES rather than Auto-Slim under the reported CIFAR-10 compression setting | MobileNetV2 · CIFAR-10 | Top-1 accuracy | Helped |
| Increase the epitome compression rate to 2.35× (ablation) | NES-1 · ESC-50 | accuracy: 73.0 -> 69.25 | Hurt |
| Increase the epitome compression rate to 3.16× (ablation) | NES-1 · ESC-50 | accuracy: 73.0 -> 65.5 | Hurt |
| Apply NES to the fully connected layer instead of the MobileNetV2-0.5 FC baseline | MobileNetV2-0.5 FC · CIFAR-10 | Top-1 accuracy | Helped |
Closest papers by meaning
- BlockSwap: Fisher-guided Block Substitution for Network Compression on a Budget (ICLR 2020 · similarity 0.87)
- Learnable Embedding Space for Efficient Neural Architecture Compression (ICLR 2019 · similarity 0.86)
- Low-Rank Compression of Neural Nets: Learning the Rank of Each Layer (CVPR 2020 · similarity 0.86)
- Few Sample Knowledge Distillation for Efficient Network Compression (CVPR 2020 · similarity 0.86)
- EPSD: Early Pruning with Self-Distillation for Efficient Model Compression (AAAI 2024 · similarity 0.85)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.85)
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.