Research map / Architectures and efficiency / Neural architecture search
BANANAS: Bayesian Optimization with Neural Architectures for 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 |
|---|---|---|---|
| Use the path encoding instead of the adjacency-matrix encoding (ablation) | Feedforward neural predictor for NASBench-101 architecture validation-accuracy prediction · NASBench-101 | prediction accuracy | Helped |
| Truncate the path encoding to 20 bits (ablation) | BANANAS · NASBench-101 | search performance: 364 bits to 20 bits | No clear effect |
| Use the full-length path encoding rather than truncating it (ablation) | BANANAS · NASBench-101 | search performance | Hurt |
| Use an ensemble of neural predictors rather than a Bayesian neural network (ablation) | Neural predictor uncertainty estimates · NASBench-101 | RMSCE | Helped |
| Use a path-encoding feedforward ensemble rather than an ensemble of GCNs (ablation) | BO + neural predictor framework · NASBench-101 | test error | Helped |
| Use independent Thompson sampling rather than the other tested acquisition functions (ablation) | Path-encoding feedforward ensemble with mutation optimization · NASBench-101 | search performance | Helped |
| Optimize acquisition by mutation rather than random sampling, local search, or evolution (ablation) | Path-encoding feedforward ensemble with BO · NASBench-101 | search performance | Helped |
| Use BANANAS rather than Local search | NAS algorithm · NASBench-101 | test error: 5.932 -> 5.923 | Helped |
| Use BANANAS rather than DARTS | NAS algorithm · DARTS search space | Avg. Test error: 2.76 -> 2.64 | Helped |
Closest papers by meaning
- Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels (ICLR 2021 · similarity 0.91)
- Neural Architecture Search with Bayesian Optimisation and Optimal Transport (NeurIPS 2018 · similarity 0.91)
- Transfer NAS with Meta-learned Bayesian Surrogates (ICLR 2023 · similarity 0.91)
- ProxyBO: Accelerating Neural Architecture Search via Bayesian Optimization with Zero-Cost Proxies (AAAI 2023 · similarity 0.91)
- Neural Architecture Generator Optimization (NeurIPS 2020 · similarity 0.91)
- Robustifying and Boosting Training-Free Neural Architecture Search (ICLR 2024 · 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.