Research map / Label-efficient and robust learning / Semi-supervised learning
A Semi-Supervised Assessor of Neural Architectures
Check 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 |
|---|---|---|---|
| use the semi-supervised assessor instead of E2EPP | NAS-Bench-101 performance prediction with 1k labeled architectures · NAS-Bench-101, 1k labeled | KTau: 0.5705 -> 0.6541 | Helped |
| use the semi-supervised assessor instead of E2EPP | NAS-Bench-101 performance prediction with 1k labeled architectures · NAS-Bench-101, 1k labeled | MSE: 0.0042 -> 0.0031 | Helped |
| use the semi-supervised assessor instead of E2EPP | NAS-Bench-101 performance prediction with 10k labeled architectures · NAS-Bench-101, 10k labeled | KTau: 0.6941 -> 0.7814 | Helped |
| use the semi-supervised assessor instead of E2EPP | NAS-Bench-101 performance prediction with 100k labeled architectures · NAS-Bench-101, 100k labeled | r: 0.5874 -> 0.8047 | Helped |
| use the semi-supervised assessor in EA search instead of E2EPP | NAS-Bench-101 architecture search with 1k labeled architectures · NAS-Bench-101, EA search | CIFAR-10 top-1 accuracy: 93.77 -> 94.01 | Helped |
| use the semi-supervised assessor in EA search instead of E2EPP | CIFAR-100 architecture search with 1k labeled architectures · CIFAR-100, EA search | top-1 accuracy: 75.86 -> 78.64 | Helped |
| use the semi-supervised assessor in EA search instead of E2EPP | CIFAR-100 architecture search with 1k labeled architectures · CIFAR-100, EA search | top-5 accuracy: 93.11 -> 94.23 | Helped |
| add an auto-encoder to learn architecture representations (ablation) | Semi-Supervised Assessor · NAS-Bench-101, 1k labeled | KTau: 0.5302 -> 0.6541 | Helped |
| add an auto-encoder to learn architecture representations (ablation) | Semi-Supervised Assessor · NAS-Bench-101, 10k labeled | KTau: 0.7188 -> 0.7814 | Helped |
| add an auto-encoder to learn architecture representations (ablation) | Semi-Supervised Assessor · NAS-Bench-101, 100k labeled | KTau: 0.7578 -> 0.8456 | Helped |
| include reconstruction loss rather than setting its weight to zero (ablation) | Semi-Supervised Assessor · NAS-Bench-101 | KTau | Helped |
| increase the number of unlabeled architectures (ablation) | Semi-Supervised Assessor · NAS-Bench-101 | KTau | Helped |
Closest papers by meaning
- Semi-Supervised Neural Architecture Search (NeurIPS 2020 · similarity 0.92)
- Contrastive Neural Architecture Search With Neural Architecture Comparators (CVPR 2021 · similarity 0.91)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.90)
- NAS-Bench-101: Towards Reproducible Neural Architecture Search (ICML 2019 · similarity 0.90)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.89)
- Insights from the Use of Previously Unseen Neural Architecture Search Datasets (CVPR 2024 · 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.