Research map / Architectures and efficiency / Neural architecture search
Rethinking Architecture Selection in Differentiable NAS
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 |
|---|---|---|---|
| Replace magnitude-based selection with perturbation-based architecture selection | DARTS · CIFAR-10 | test error (%): 3.00% to 2.61% | Helped |
| Replace magnitude-based operation selection with perturbation-based selection | SDARTS-RS · CIFAR-10 | test error (%): 2.67% to 2.54% | Helped |
| Replace magnitude-based operation selection with perturbation-based selection | SGAS · CIFAR-10 | test error (%): 2.66% to 2.56% | Helped |
| Replace magnitude-based selection with perturbation-based architecture selection | DARTS · ImageNet mobile setting | top-1 test error (%): 26.7 to 25.5 | Helped |
| Replace magnitude-based selection with perturbation-based architecture selection | DARTS · ImageNet mobile setting | top-5 test error (%): 8.7 to 8.0 | Helped |
| Use magnitude-based operation selection instead of perturbation-based selection (ablation) | DARTS+PT · CIFAR-10, DARTS search space | test error (%): 2.61% to 2.85% | Hurt |
| Fix α=0 during supernet training (ablation) | DARTS+PT · CIFAR-10, DARTS search space | test error (%): 2.61 to 2.87 | Hurt |
| Fix α=0 during supernet training (ablation) | DARTS+PT · CIFAR-10, NAS-Bench-201 | test error (%): 11.89 to 6.20 | Helped |
| Increase fine-tuning beyond 5 epochs between edge decisions (ablation) | DARTS+PT · DARTS search space | test error | No clear effect |
Closest papers by meaning
- iDARTS: Differentiable Architecture Search with Stochastic Implicit Gradients (ICML 2021 · similarity 0.93)
- BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule (CVPR 2022 · similarity 0.93)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.92)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.91)
- Boosting Order-Preserving and Transferability for Neural Architecture Search: a Joint Architecture Refined Search and Fine-tuning Approach (CVPR 2024 · similarity 0.91)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.90)
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.