Research map / Architectures and efficiency / Neural architecture search
Cream of the Crop: Distilling Prioritized Paths For One-Shot Neural Architecture Search
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 distillation from a meta-matched prioritized path (ablation) | single-path one-shot NAS with prioritized path board · ImageNet val | top-1 accuracy (%): 76.5 -> 79.2 | Helped |
| replace random matching with meta-learned matching of prioritized paths (ablation) | prioritized path distillation · ImageNet val | top-1 accuracy (%): 77.9 -> 79.2 | Helped |
| replace fixed largest-model teacher with meta-learned prioritized path teacher (ablation) | prioritized path distillation · ImageNet val | top-1 accuracy (%): 77.6 -> 79.2 | Helped |
| distill from random prioritized path versus no distillation (ablation) | prioritized path board · ImageNet val | top-1 accuracy (%): 76.5 -> 77.9 | Helped |
| replace evolution search with prioritized path selection (no distillation) (ablation) | single-path one-shot NAS · ImageNet val | top-1 accuracy (%): 76.3 -> 76.5 | No clear effect |
| add evolution search on top of distilled hypernetwork (ablation) | Cream with meta match · ImageNet val | top-1 accuracy (%): 79.2 -> 79.2 | No clear effect |
| add prioritized path distillation with meta matching (ablation) | single-path one-shot NAS · subImageNet | Kendall rank correlation: 0.19 -> 0.37 | Helped |
| use Cream instead of SPOS in a harder search space (MBConv + ResBlock + 2D Conv) | one-shot NAS · ImageNet 120 epochs | top-1 accuracy (%): 74.0 -> 77.1 | Helped |
| increase number of prioritized paths (ablation) | Cream · hypernetwork performance | hypernetwork accuracy | Helped |
Closest papers by meaning
- Distribution Consistent Neural Architecture Search (CVPR 2022 · similarity 0.91)
- Can Weight Sharing Outperform Random Architecture Search? An Investigation With TuNAS (CVPR 2020 · similarity 0.90)
- AlphaNet: Improved Training of Supernets with Alpha-Divergence (ICML 2021 · similarity 0.89)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.88)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.88)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.88)
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.