Research map / Architectures and efficiency / Neural architecture search
Searching for a Robust Neural Architecture in Four GPU Hours
Check 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 |
|---|---|---|---|
| fix the reduction cell and search only the normal cell (ablation) | GDAS · CIFAR-10 test | error (%): 2.93 -> 2.82 | Helped |
| fix the reduction cell while using the GDAS normal cell (ablation) | GDAS · CIFAR-10 test | error (%): 2.93 -> 2.87 | Helped |
| use the FRC normal cell while searching the reduction cell (ablation) | GDAS · CIFAR-10 test | error (%): 2.93 -> 2.84 | Helped |
| add CutOut | GDAS (FRC) · CIFAR-10 test | error (%): 3.75 -> 2.82 | Helped |
| add CutOut | GDAS · CIFAR-10 test | error (%): 3.87 -> 2.93 | Helped |
| use the fixed reduction cell rather than searching the reduction cell (ablation) | GDAS · ImageNet validation | Top-1 error (%): 26.0 -> 27.5 | Hurt |
| transfer the PTB-searched architecture to WT2 | GDAS · WT2 test | perplexity: 69.4 vs 65.9 | Hurt |
| apply the acceleration step that computes only the sampled operation | GDAS · CNN architecture search | search computation per iteration: 8× | Helped |
Closest papers by meaning
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.90)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.90)
- ISTA-NAS: Efficient and Consistent Neural Architecture Search by Sparse Coding (NeurIPS 2020 · similarity 0.89)
- UNAS: Differentiable Architecture Search Meets Reinforcement Learning (CVPR 2020 · similarity 0.88)
- CE-NAS: An End-to-End Carbon-Efficient Neural Architecture Search Framework (NeurIPS 2024 · similarity 0.88)
- ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware (ICLR 2019 · 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.