Research map / Architectures and efficiency / Neural architecture search
NetAdaptV2: Efficient Neural Architecture Search With Fast Super-Network Training and Architecture Optimization
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 NetAdaptV2 rather than MobileNetV3 | ImageNet latency-guided search · ImageNet | Top-1 accuracy (%): 75.2 -> 77.0 | Helped |
| Use NetAdaptV2 rather than NetAdaptV1 | NYU Depth V2 search · NYU Depth V2 | Delta-1 accuracy (%): 77.4 -> 77.9 | Helped |
| Use NetAdaptV2 rather than NetAdaptV1 | NYU Depth V2 search · NYU Depth V2 | Search time (GPU-hours, NYU Depth): 65 -> 27; 2.4× | Helped |
| Add ordered dropout (ablation) | NetAdaptV2 super-network training · ImageNet, MobileNetV1 ablation | Search time (GPU-hours): 721 -> 221; 3.3× | Helped |
| Add ordered dropout (ablation) | NetAdaptV2 super-network training · ImageNet, MobileNetV1 ablation | Top-1 accuracy (%): 71.0 -> 71.1 (+0.1) | No clear effect |
| Add channel-level bypass connections (ablation) | NetAdaptV2 · ImageNet, latency around 51ms | Top-1 accuracy (%): 75.9 -> 76.2 (+0.3) | Helped |
| Add multi-layer coordinate descent (ablation) | NetAdaptV2 with CBC · ImageNet, latency around 51ms | Top-1 accuracy (%): 76.2 -> 76.6 (+0.4) | Helped |
| Set initial latency reduction to 6.0% rather than 3.0% (ablation) | MCD optimizer · ImageNet, fixed samples per iteration | Top-1 accuracy (%): 76.4 -> 75.9 | Hurt |
| Increase samples from 100 to 200 while setting latency reduction to 3.0% (ablation) | MCD optimizer · ImageNet, fixed evaluation time | Top-1 accuracy (%): 76.4 -> 76.6 | Helped |
| Increase samples to 400 while setting latency reduction to 6.0% (ablation) | MCD optimizer · ImageNet, fixed evaluation time | Top-1 accuracy (%): 76.4 -> 75.7 | Hurt |
| Add channel-level bypass connections (ablation) | NetAdaptV2 on MobileNetV1 · ImageNet, latency around 6.5ms | Top-1 accuracy (%): 40.0 -> 46.5 (+6.5) | Helped |
| Add multi-layer coordinate descent to CBC (ablation) | NetAdaptV2 on MobileNetV1 · ImageNet, latency around 6.5ms | Top-1 accuracy (%): 46.5 -> 49.3 (+2.8) | Helped |
Closest papers by meaning
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.93)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.91)
- Distribution Consistent Neural Architecture Search (CVPR 2022 · similarity 0.91)
- Boosting Order-Preserving and Transferability for Neural Architecture Search: a Joint Architecture Refined Search and Fine-tuning Approach (CVPR 2024 · similarity 0.91)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.90)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.90)
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.