Research map / Architectures and efficiency / Neural architecture search
Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets
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 |
|---|---|---|---|
| Generate dataset-conditioned architectures and select them with the meta-performance predictor rather than random sampling (ablation) | MetaD2A · Aircraft | accuracy (%): 38.15 ±0.99 -> 58.43 ±1.18 | Helped |
| Generate dataset-conditioned architectures and select them with the meta-performance predictor rather than random sampling (ablation) | MetaD2A · CIFAR-10 | accuracy (%): 93.06 ±0.55 -> 94.37 ±0.03 | Helped |
| Generate dataset-conditioned architectures and select them with the meta-performance predictor rather than random sampling (ablation) | MetaD2A · CIFAR-100 | accuracy (%): 69.94 ±1.21 -> 73.51 ±0.00 | Helped |
| Add the meta-performance predictor to the generator (ablation) | MetaD2A · CIFAR-10 | accuracy (%): 93.96 ±0.22 -> 94.37 ±0.03 | Helped |
| Add the meta-performance predictor to the generator (ablation) | MetaD2A · CIFAR-100 | accuracy (%): 71.54 ±0.63 -> 73.51 ±0.00 | Helped |
| Add the meta-performance predictor to the generator (ablation) | MetaD2A · Aircraft | accuracy (%): 53.45 ±3.27 -> 58.43 ±1.18 | Helped |
| Add dataset input to the graph-only performance predictor (ablation) | MetaD2A predictor · NAS-Bench-201 validation tasks | Pearson correlation coefficient: 0.6439 -> 0.8085 | Helped |
| Use the proposed class-aware set encoder instead of DeepSet (ablation) | MetaD2A predictor · NAS-Bench-201 validation tasks | Pearson correlation coefficient: 0.7286 -> 0.8085 | Helped |
| Use the proposed class-aware set encoder instead of SetTransformer (ablation) | MetaD2A predictor · NAS-Bench-201 validation tasks | Pearson correlation coefficient: 0.7744 -> 0.8085 | Helped |
| Use the proposed class-aware set encoder instead of Statistical Pooling (ablation) | MetaD2A predictor · NAS-Bench-201 validation tasks | Pearson correlation coefficient: 0.7796 -> 0.8085 | Helped |
| Use MetaD2A rather than DrNAS for architecture search | NAS-Bench-201 NAS · CIFAR-100 | accuracy (%): 73.51 ±0.00 -> 73.51 ±0.00 | No clear effect |
| Use MetaD2A rather than DrNAS for architecture search | NAS-Bench-201 NAS · CIFAR-10 | accuracy (%): 94.36 ±0.00 -> 94.37 ±0.03 | Helped |
Closest papers by meaning
- Meta Architecture Search (NeurIPS 2019 · similarity 0.92)
- M-NAS: Meta Neural Architecture Search (AAAI 2020 · similarity 0.91)
- Rapid Model Architecture Adaption for Meta-Learning (NeurIPS 2022 · similarity 0.91)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.91)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 2024 · similarity 0.90)
- 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.