Research map / Architectures and efficiency / Neural architecture search
Single Path One-Shot Neural Architecture Search with Uniform Sampling
Zichao Guo, Xiangyu Zhang, Haoyuan Mu, Wen Heng, Zechun Liu, Yichen Wei, Jian Sun
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 evolutionary architecture search instead of random search on the single path supernet (ablation) | SPOS · ImageNet block search | top-1 acc (%): 73.8 -> 74.3 | Improved |
| search building blocks instead of fixed all choice_3 | SPOS · ImageNet | top-1 acc (%): 73.4 -> 74.3 | Improved |
| add channel search on all choice_3 | SPOS · ImageNet | top-1 acc (%): 73.4 -> 73.9 | Improved |
| add channel search on top of block search | SPOS · ImageNet | top-1 acc (%): 74.3 -> 74.7 | Improved |
| search channels instead of random channel selection (ablation) | SPOS · ImageNet | top-1 acc (%): 73.1 -> 73.9 | Improved |
| run SPOS at same latency as FBNet-B | SPOS · ImageNet | top-1 acc (%): 74.1 -> 74.8 | Improved |
| run SPOS at same FLOPs as FBNet-A | SPOS · ImageNet | top-1 acc (%): 73.0 -> 73.2 | Improved |
| run SPOS at same latency as Proxyless (GPU) | SPOS · ImageNet | top-1 acc (%): 74.7 -> 75.3 | Improved |
| search mixed-precision bit widths instead of uniform 2W2A | SPOS · ResNet-18 quantization | top-1 acc (%): 65.6 -> 66.4 | Improved |
| search mixed-precision instead of uniform 3W3A | SPOS · ResNet-34 quantization | top-1 acc (%): 72.5 -> 73.9 | Improved |
| search mixed precision instead of DNAS at 4W4A | SPOS · ResNet-18 quantization | top-1 acc (%): 70.6 -> 70.5 | Mixed |
| reduce NAS-Bench-201 search space by deleting both 1x1 conv and avg pooling (Reduce-3) (ablation) | SPOS supernet · NAS-Bench-201 CIFAR-10 | Kendall tau: 0.55 -> 0.64 | Improved |
Closest papers by meaning
- GreedyNAS: Towards Fast One-Shot NAS With Greedy Supernet (CVPR 2020 · similarity 0.93)
- Few-Shot Neural Architecture Search (ICML 2021 · similarity 0.92)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.91)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.90)
- Revisiting Neural Networks for Few-Shot Learning: A Zero-Cost NAS Perspective (ICML 2025 · similarity 0.89)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.89)
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.
