Research map / Architectures and efficiency / Neural architecture search
Network Adjustment: Channel Search Guided by FLOPs Utilization Ratio
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 |
|---|---|---|---|
| Adjust channel widths using FUR-guided reallocations at approximately fixed FLOPs | ResNet-20 · CIFAR-100 | accuracy: 69.52 -> 71.57 | Helped |
| Adjust channel widths using FUR-guided reallocations at approximately fixed FLOPs | ResNet-32 · CIFAR-100 | accuracy: 72.57 -> 74.14 | Helped |
| Adjust channel widths using FUR-guided reallocations at approximately fixed FLOPs | ResNet-110 · CIFAR-100 | accuracy: 75.68 -> 76.98 | Helped |
| Adjust channel widths with Network Adjustment | ResNet-18 · ImageNet | Top-1 accuracy: 70.31 -> 71.22 | Helped |
| Adjust channel widths with Network Adjustment | MnasNet · ImageNet | Top-1 accuracy: 74.21 -> 74.88 | Helped |
| Add FUR-based FLOPs-aware layer ranking instead of fixed-ratio channel-drop ranking (ablation) | Network Adjustment on ResNet-20 · CIFAR-100 | accuracy: 71.06% -> 71.57% | Helped |
| Add channel adjustment to the initial ResNet-20 channel configuration | ResNet-20 · CIFAR-100 | accuracy: 69.52 -> 71.57 | Helped |
| Transfer the searched ResNet-20 channel pattern by doubling channels rather than searching directly (ablation) | ResNet-20 2.0× · CIFAR-100 | accuracy: 76.34% vs 76.87% | Hurt |
| Use knowledge distillation with the adjusted ResNet-18 0.8× | ResNet-18 0.8× · ImageNet | Top-1 accuracy: 69.41 -> 71.00 | Helped |
| Use a poor PyramidNet-20 channel initialization, then search channel widths (ablation) | PyramidNet-20 · CIFAR-100 | accuracy: 66.58 -> 71.53 | Helped |
Closest papers by meaning
- AOWS: Adaptive and Optimal Network Width Search With Latency Constraints (CVPR 2020 · similarity 0.86)
- Revisiting Random Channel Pruning for Neural Network Compression (CVPR 2022 · similarity 0.85)
- Network Pruning via Performance Maximization (CVPR 2021 · similarity 0.84)
- Run, Don't Walk: Chasing Higher FLOPS for Faster Neural Networks (CVPR 2023 · similarity 0.84)
- BCNet: Searching for Network Width With Bilaterally Coupled Network (CVPR 2021 · similarity 0.84)
- How Sparse Can We Prune A Deep Network: A Fundamental Limit Perspective (NeurIPS 2024 · similarity 0.83)
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.