Research map / Architectures and efficiency / Neural architecture search
How Powerful are Performance Predictors in Neural Architecture Search?
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 |
|---|---|---|---|
| combine learning-curve (SoTL-E) and zero-cost (Jacobian covariance) features with a model-based predictor (OMNI) | SemiNAS/NGBoost predictors · NAS-Bench search spaces | Kendall Tau: up to 30% higher | Helped |
| add zero-cost proxy features to NGBoost in predictor-guided NAS | NGBoost · NAS-Bench-201 CIFAR-10/ImageNet16-120 | validation error vs runtime | Helped |
| use zero-cost predictor (Jacobian covariance/SynFlow) instead of model-based predictors after 30 hours | model-based predictors · NAS-Bench-101/201 | Kendall Tau | Helped |
| use zero-cost predictors on larger DARTS space | zero-cost proxies · DARTS | Kendall Tau | Hurt |
| mutation-based test set instead of uniform random test set | model-based predictors · NAS-Bench-201 CIFAR-10 | Kendall Tau | Hurt |
| use path encoding with tree/GP predictors instead of adjacency encoding | simple tree-based and GP predictors · NAS-Bench-101 | Kendall Tau | Helped |
Closest papers by meaning
- Dynamic Ensemble of Low-Fidelity Experts: Mitigating NAS “Cold-Start” (AAAI 2023 · similarity 0.92)
- Speedy Performance Estimation for Neural Architecture Search (NeurIPS 2021 · similarity 0.92)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.91)
- Stronger NAS with Weaker Predictors (NeurIPS 2021 · similarity 0.91)
- Contrastive Neural Architecture Search With Neural Architecture Comparators (CVPR 2021 · similarity 0.91)
- 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.