Research map / Architectures and efficiency / Neural architecture search
Beyond the Limits: Overcoming Negative Correlation of Activation-Based Training-Free NAS
Haidong Kang, Lianbo Ma, Pengjun Chen, Guo Yu, Xingwei Wang, Min Huang
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 |
|---|---|---|---|
| add stochastic activation masking (SAM) to SWAP (ablation) | SWAP · NAS-Bench-201 ImageNet16-120 | test accuracy (%): 35.40 -> 42.48 | Improved |
| add non-linear rescaling (NIR) to SWAP (ablation) | SWAP · NAS-Bench-201 ImageNet16-120 | test accuracy (%): 35.40 -> 44.49 | Improved |
| add both SAM and NIR to SWAP (ablation) | SWAP · NAS-Bench-201 CIFAR-10 | test accuracy (%): 90.48 -> 93.68 | Improved |
| add both SAM and NIR to NWOT (ablation) | NWOT · NAS-Bench-201 ImageNet16-120 | test accuracy (%): 42.31 -> 45.27 | Improved |
| add SAM only to NWOT (ablation) | NWOT · NAS-Bench-201 ImageNet16-120 | test accuracy (%): 42.31 -> 42.93 | No clear effect |
| apply NCD to NWOT in the 77-convolution NAS-Bench-201 subspace | NWOT · NAS-Bench-201 subspace 77 (CIFAR-10) | Spearman rho: -0.734 -> 0.679 | Improved |
| apply NCD to SWAP in the 77-convolution subspace | SWAP · NAS-Bench-201 subspace 77 (CIFAR-10) | Spearman rho: -0.246 -> 0.715 | Improved |
| apply NCD in the 5-convolution subspace | SWAP and NWOT · NAS-Bench-201 subspace 5 (CIFAR-10) | Spearman rho: 0.769 -> 0.765 (SWAP); 0.773 -> 0.747 (NWOT) | Hurt |
| apply NCD to SWAP on NAS-Bench-201 CIFAR-10 | SWAP · NAS-Bench-201 CIFAR-10 | test accuracy (%): 90.48 -> 93.68 | Improved |
| vary SAM masking probability alpha (ablation) | SWAP+SAM · NAS-Bench-201 CIFAR-10 | test accuracy (%): 90.48 (alpha=0) -> 93.15 (alpha=0.95) -> 93.05 (alpha=0.99) | Mixed |
| vary alpha for NWOT+SAM+NIR (ablation) | NWOT+SAM+NIR · NAS-Bench-201 CIFAR-10 | test accuracy (%): 93.48 (alpha=0) -> 94.00 (alpha=0.95) -> 93.30 (alpha=0.99) | Mixed |
| apply NCD-NWOT to MobileNet-like space search | NCD-NWOT · MobileNet-like ImageNet1k | top-1 accuracy (%): 75.6 -> 78.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- ZiCo — ZiCo: Zero-shot NAS via inverse Coefficient of Variation on Gradients (ICLR 2023) · better on 2 of 2 shared results
- AZ-NAS — AZ-NAS: Assembling Zero-Cost Proxies for Network Architecture Search (CVPR 2024) · better on 1 of 1 shared result
- DARTS+ — DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021) · better on 1 of 1 shared result
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
- Unifying and Boosting Gradient-Based Training-Free Neural Architecture Search (NeurIPS 2022 · similarity 0.90)
- Understanding and Accelerating Neural Architecture Search With Training-Free and Theory-Grounded Metrics (TPAMI 2023 · similarity 0.89)
- SWAP-NAS: Sample-Wise Activation Patterns for Ultra-fast NAS (ICLR 2024 · similarity 0.88)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.88)
- Robustifying and Boosting Training-Free Neural Architecture Search (ICLR 2024 · similarity 0.88)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.87)
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.
