Research map / Architectures and efficiency / Neural architecture search
An Analysis of Super-Net Heuristics in Weight-Sharing NAS
Kaicheng Yu, René Ranftl, Mathieu Salzmann
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 batch statistics at test time instead of tracked running statistics (affine false) (ablation) | Random-NAS super-net · NASBench-201 | sparse Kendall-Tau: -0.171 -> 0.783 | Improved |
| disable dynamic channeling in training and test (ablation) | Random-NAS super-net · NASBench-101 | sparse Kendall-Tau: 0.22 -> 0.46 | Improved |
| disable dynamic channeling in training and test (ablation) | Random-NAS super-net · NASBench-101 | stand-alone accuracy: 89.43 -> 93.65 | Improved |
| disable dynamic channeling only at test time (ablation) | Random-NAS super-net · NASBench-101 | stand-alone accuracy: 89.43 -> 91.27 | Improved |
| use a lower learning rate 0.025 instead of 0.1 (ablation) | Random-NAS super-net · NASBench-201 | stand-alone accuracy: 91.82 -> 92.42 | Improved |
| train for more epochs beyond 400 (ablation) | Random-NAS super-net · NASBench-101, NASBench-201, DARTS-NDS | sparse Kendall-Tau / final accuracy | No clear effect |
| use weight decay 0.0003 instead of none (ablation) | Random-NAS super-net · NASBench-101 | sparse Kendall-Tau: -0.037 -> 0.144 | Mixed |
| reduce repeated cells from 3 to 1 (ablation) | Random-NAS super-net · NASBench-201 | sparse Kendall-Tau: 0.751 -> 0.502 | Hurt |
| reduce batch size from 256 to 64 (ablation) | Random-NAS super-net · NASBench-201 | sparse Kendall-Tau: 0.740 -> 0.703 | Hurt |
| reduce training portion to 0.25 (ablation) | Random-NAS super-net · NASBench-201 | sparse Kendall-Tau: 0.751 -> 0.693 | Mixed |
| reduce initial channels from 16 to 4 (ablation) | Random-NAS super-net · NASBench-201 | sparse Kendall-Tau: 0.740 -> 0.691 | No clear effect |
| add weight-sharing batch norm (WSBN) (ablation) | Random-NAS super-net · NASBench-101 | sparse Kendall-Tau: 0.236 -> 0.056 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DARTS-V2 — DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Distribution Consistent Neural Architecture Search (CVPR 2022 · similarity 0.93)
- Landmark Regularization: Ranking Guided Super-Net Training in Neural Architecture Search (CVPR 2021 · similarity 0.92)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.91)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · 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 0.91)
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.
