Research map / Architectures and efficiency / Neural architecture search
Efficient Neural Architecture Search via Parameters Sharing
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 ENAS-discovered recurrent architecture instead of NAS | Penn Treebank language modeling · Penn Treebank test | perplexity: 62.4 -> 55.8 | Helped |
| add CutOut (ablation) | ENAS micro · CIFAR-10 test | error (%): 3.54 -> 2.89 | Helped |
| increase the highest layer to 512 filters (ablation) | ENAS macro · CIFAR-10 test | error (%): 4.23 -> 3.87 | Helped |
| replace separable convolutions with normal convolutions while matching parameter count (ablation) | ENAS macro architecture · CIFAR-10 test | error (%): +1.7% | Hurt |
| randomly change several connections (ablation) | ENAS micro cells · CIFAR-10 test | error (%): +2.1% | Hurt |
| replace selected recurrent-cell nonlinearities with identity or sigmoid (ablation) | ENAS recurrent cell · Penn Treebank test | perplexity: up to 8 points | Hurt |
| swap selected ReLU nonlinearities with tanh or vice versa (ablation) | ENAS recurrent cell · Penn Treebank test | perplexity: up to 3 points | Hurt |
| replace controller-guided architecture selection with uniform random sampling (ablation) | ENAS search spaces · Penn Treebank and CIFAR-10 | perplexity / error (%): 55.8 vs 81.2; 4.23 vs 5.86; 3.54 vs 6.77 | Hurt |
| disable controller updates during shared-weight training (ablation) | ENAS macro · CIFAR-10 test | error (%): 4.23 -> 8.92 | Hurt |
Closest papers by meaning
- Speedy Performance Estimation for Neural Architecture Search (NeurIPS 2021 · similarity 0.89)
- Bridging the Gap between Sample-based and One-shot Neural Architecture Search with BONAS (NeurIPS 2020 · similarity 0.88)
- Neural Architecture Optimization (NeurIPS 2018 · similarity 0.88)
- Neural Parameter Allocation Search (ICLR 2022 · similarity 0.88)
- Runtime Analysis of Evolutionary NAS for Multiclass Classification (ICML 2025 · similarity 0.88)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.87)
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.