Research map / Architectures and efficiency / Neural architecture search
AtomNAS: Fine-Grained End-to-End 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 FLOPs-aware coefficients in the atomic-block penalty rather than equal penalties (ablation) | AtomNAS · ImageNet | Top-1 accuracy (%): 74.9 -> 75.9 | Helped |
| Use FLOPs-aware coefficients in the atomic-block penalty rather than equal penalties (ablation) | AtomNAS · ImageNet | FLOPs: 445M -> 360M | Mixed |
| Recalculate BN running statistics before inference (ablation) | AtomNAS-A · ImageNet | Top-1 accuracy (%): 73.2 -> 74.6 (+1.4) | Helped |
| Recalculate BN running statistics before inference (ablation) | AtomNAS-B · ImageNet | Top-1 accuracy (%): 73.8 -> 75.5 (+1.7) | Helped |
| Recalculate BN running statistics before inference (ablation) | AtomNAS-C · ImageNet | Top-1 accuracy (%): 74.7 -> 75.9 (+1.2) | Helped |
| Replace ReLU with Swish and add SE modules with ratio 0.5 | AtomNAS-A · ImageNet | Top-1 accuracy (%): 74.6 -> 76.3 | Helped |
| Replace ReLU with Swish and add SE modules with ratio 0.5 | AtomNAS-B · ImageNet | Top-1 accuracy (%): 75.5 -> 77.2 | Helped |
| Replace ReLU with Swish and add SE modules with ratio 0.5 | AtomNAS-C · ImageNet | Top-1 accuracy (%): 75.9 -> 77.6 | Helped |
| Use dynamic network shrinkage to remove dead atomic blocks during training | AtomNAS-C · ImageNet search and training | Training cost increase vs. training the searched model from scratch (%): 17.2% | Helped |
| Use mixed atomic-block operations in the searched backbone | AtomNAS · COCO val2017 | detect-mAP (%): 32.0 -> 34.1 | Helped |
Closest papers by meaning
- ISTA-NAS: Efficient and Consistent Neural Architecture Search by Sparse Coding (NeurIPS 2020 · similarity 0.92)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.92)
- Understanding Architectures Learnt by Cell-based Neural Architecture Search (ICLR 2020 · similarity 0.91)
- einspace: Searching for Neural Architectures from Fundamental Operations (NeurIPS 2024 · similarity 0.91)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.90)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 2024 · similarity 0.89)
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.