Research map / Architectures and efficiency / Neural architecture search
ProxyBO: Accelerating Neural Architecture Search via Bayesian Optimization with Zero-Cost Proxies
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 |
|---|---|---|---|
| Add dynamically weighted zero-cost proxy rankings to BO | PRF-based NAS search · NAS-Bench-101 | test error (%): 6.14 -> 5.96 | Helped |
| Add dynamically weighted zero-cost proxy rankings to BO | PRF-based NAS search · NAS-Bench-201 CIFAR-10 | test error (%): 8.80 -> 8.54 | Helped |
| Add dynamically weighted zero-cost proxy rankings to BO | PRF-based NAS search · NAS-Bench-201 CIFAR-100 | test error (%): 27.03 -> 26.52 | Helped |
| Add dynamically weighted zero-cost proxy rankings to BO | PRF-based NAS search · NAS-Bench-201 ImageNet16-120 | test error (%): 53.12 -> 52.82 | Helped |
| Add dynamically weighted zero-cost proxy rankings to BO | PRF-based NAS search · NAS-Bench-ASR | test PER (%): 21.47±0.06 -> 21.43±0.03 | Helped |
| Apply ProxyBO to GPWL instead of using GPWL alone | GPWL-based NAS search · NAS-Bench-101 | test error (%): 5.88 -> 5.83 | Helped |
| Apply ProxyBO to GPWL instead of using GPWL alone | GPWL-based NAS search · NAS-Bench-201 CIFAR-10 | test error (%): 8.56 -> 8.52 | Helped |
| Apply ProxyBO to GPWL instead of using GPWL alone | GPWL-based NAS search · NAS-Bench-201 CIFAR-100 | test error (%): 26.55 -> 26.50 | Helped |
| Apply ProxyBO to GPWL instead of using GPWL alone | GPWL-based NAS search · NAS-Bench-201 ImageNet16-120 | test error (%): 52.97 -> 52.78 | Helped |
| Use synflow rather than all three proxies (ablation) | ProxyBO with PRF · NAS-Bench-101 | test error (%): 5.96 -> 5.87 | Helped |
| Use synflow rather than all three proxies (ablation) | ProxyBO with PRF · NAS-Bench-201 ImageNet16-120 | test error (%): 52.82 -> 52.74 | Helped |
| Use snip proxy (ablation) | ProxyBO with PRF · NAS-Bench-101 | test error (%): 6.14 -> 6.18 | Hurt |
Closest papers by meaning
- EZNAS: Evolving Zero-Cost Proxies For Neural Architecture Scoring (NeurIPS 2022 · similarity 0.92)
- EcoNAS: Finding Proxies for Economical Neural Architecture Search (CVPR 2020 · similarity 0.92)
- Speedy Performance Estimation for Neural Architecture Search (NeurIPS 2021 · similarity 0.91)
- BANANAS: Bayesian Optimization with Neural Architectures for Neural Architecture Search (AAAI 2021 · similarity 0.91)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.89)
- ParZC: Parametric Zero-Cost Proxies for Efficient NAS (AAAI 2025 · similarity 0.88)
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.