Research map / Architectures and efficiency / Neural architecture search
NAS-Bench-Suite: NAS Evaluation is (Now) Surprisingly Easy
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 |
|---|---|---|---|
| tune hyperparameters separately on each benchmark | Regularized Evolution · 25 NAS benchmark settings | average rank: 2.36 -> 1.96 | Helped |
| tune hyperparameters separately on each benchmark | XGBoost · 25 NAS benchmark settings | average rank: 2.95 -> 1.23 | Helped |
| transfer tuned hyperparameters from another benchmark rather than tuning on the target | NAS algorithms and performance predictors · Other NAS benchmarks | performance | Hurt |
| use performance predictors tuned on NAS-Bench-101 and NAS-Bench-201 for other benchmarks | Performance predictors · NAS benchmarks outside NAS-Bench-101 and NAS-Bench-201 | average rank | Mixed |
| use Regularized Evolution instead of random search | Black-box NAS algorithms · 25 NAS benchmark settings | relative performance | Mixed |
| choose a single one-shot algorithm across all evaluated benchmarks | DARTS, GDAS, and DrNAS · Nine NAS benchmarks | relative performance | Mixed |
Closest papers by meaning
- NAS-Bench-360: Benchmarking Neural Architecture Search on Diverse Tasks (NeurIPS 2022 · similarity 0.95)
- NAS-Bench-201: Extending the Scope of Reproducible Neural Architecture Search (ICLR 2020 · similarity 0.94)
- Surrogate NAS Benchmarks: Going Beyond the Limited Search Spaces of Tabular NAS Benchmarks (ICLR 2022 · similarity 0.93)
- NAS-Bench-101: Towards Reproducible Neural Architecture Search (ICML 2019 · similarity 0.92)
- NAS evaluation is frustratingly hard (ICLR 2020 · similarity 0.92)
- Insights from the Use of Previously Unseen Neural Architecture Search Datasets (CVPR 2024 · similarity 0.90)
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.