Research map / Architectures and efficiency / Neural architecture search
Zero-Shot Neural Network Evaluation With Sample-Wise Activation Patterns
Yameng Peng, Andy Song, Haytham M. Fayek, Vic Ciesielski, Xiaojun Chang
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 Gaussian size regularisation to SWAP-Score (ablation) | SWAP-Score · NB101 CIFAR-10 | Spearman correlation: 0.47 -> 0.75 | Improved |
| add Gaussian size regularisation to SWAP-Score (ablation) | SWAP-Score · NB201 CIFAR-100 | Spearman correlation: 0.81 -> 0.90 | Improved |
| add Gaussian size regularisation to SWAP-Score (ablation) | SWAP-Score · NB301 CIFAR-10 | Spearman correlation: 0.58 -> 0.63 | Improved |
| add the same regularisation to #Params (ablation) | #Params · NB101 CIFAR-10 | Spearman correlation: 0.38 -> 0.76 | Improved |
| add the same regularisation to #Params (ablation) | #Params · NB201 CIFAR-100 | Spearman correlation: 0.73 -> 0.73 | No clear effect |
| add the same regularisation to #FLOPs (ablation) | #FLOPs · NB101 CIFAR-10 | Spearman correlation: 0.37 -> 0.76 | Improved |
| increase batch size from 8 to 128 (ablation) | SWAP-Score · 1000 DARTS CNNs CIFAR-10 | Spearman correlation: 0.933 -> 0.649 | Hurt |
| increase batch size (ablation) | SWAP-Score · 500 BERT-like Transformers GLUE | Spearman correlation: approximately 0.71 | No clear effect |
| increase input image dimension from 3x3x3 to 3x32x32 (ablation) | SWAP-Score · 1000 DARTS CNNs CIFAR-10 | Spearman correlation | Improved |
| replace CIFAR-10 images with Gaussian noise and increase dimension (ablation) | SWAP-Score · 1000 DARTS CNNs CIFAR-10 | Spearman correlation | Hurt |
| use SWAP-Score instead of Attention Confidence for Transformers | zero-shot proxy · GLUE 500 Transformers | Spearman correlation: 0.666 -> 0.711 | Improved |
| use SWAP-NAS adaptive on ImageNet instead of mu=25 sigma=25 | SWAP-NAS · ImageNet | top-1 test error: 24.0 -> 23.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- MeCo — MeCo: Zero-Shot NAS with One Data and Single Forward Pass via Minimum Eigenvalue of Correlation (NeurIPS 2023) · better on 2 of 3 shared results
- ZiCo — ZiCo: Zero-shot NAS via inverse Coefficient of Variation on Gradients (ICLR 2023) · better on 0 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
- SWAP-NAS: Sample-Wise Activation Patterns for Ultra-fast NAS (ICLR 2024 · similarity 0.92)
- ZiCo: Zero-shot NAS via inverse Coefficient of Variation on Gradients (ICLR 2023 · similarity 0.88)
- ParZC: Parametric Zero-Cost Proxies for Efficient NAS (AAAI 2025 · similarity 0.87)
- L-SWAG: Layer-Sample Wise Activation with Gradients Information for Zero-Shot NAS on Vision Transformers (CVPR 2025 · similarity 0.86)
- NEAR: A Training-Free Pre-Estimator of Machine Learning Model Performance (ICLR 2025 · similarity 0.86)
- Surprisingly Strong Performance Prediction with Neural Graph Features (ICML 2024 · similarity 0.84)
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.
