Research map / Architectures and efficiency / Neural architecture search
Densely Connected Search Space for More Flexible Neural Architecture Search
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 |
|---|---|---|---|
| Use the densely connected search space to automatically search block counts and widths | MobileNetV2-based ImageNet architectures · ImageNet | Top-1 accuracy: 74.3 -> 75.3 | Helped |
| Use the densely connected search space rather than the manually fixed ResNet-18 structure | ResNet-18 · ImageNet | Top-1 Acc(%): 72.0 -> 73.5 | Helped |
| Use the densely connected search space rather than the manually fixed ResNet-34 structure | ResNet-34 · ImageNet | Top-1 Acc(%): 75.3 -> 75.8 | Helped |
| Use the densely connected search space rather than the manually fixed ResNet-50-B structure | ResNet-50-B · ImageNet | FLOPs: 4.09B -> 3.41B | Helped |
| Use DenseNAS-C as the RetinaNet backbone | RetinaNet · COCO RetinaNet | mAP(%): 35 -> 35.1 | Helped |
| Use DenseNAS-C as the SSDLite backbone | SSDLite · COCO SSDLite | mAP(%): 22.9 -> 23.1 | Helped |
| Set maximum routing-block connections to 3 instead of 4 (ablation) | DenseNAS · ImageNet | Top-1 Acc(%): 75.3 -> 74.8 | Hurt |
| Increase maximum routing-block connections from 4 to 5 while keeping search at 150 epochs (ablation) | DenseNAS · ImageNet | Top-1 Acc(%): 75.3 -> 74.6 | Hurt |
| Increase search epochs to 200 with maximum routing-block connections set to 5 (ablation) | DenseNAS · ImageNet | Top-1 Acc(%): 75.3 -> 74.9 | Hurt |
| Replace chained cost estimation with local cost estimation (ablation) | DenseNAS · ImageNet | Top-1 Acc(%): 75.3 -> 74.8 | Hurt |
Closest papers by meaning
- MathNAS: If Blocks Have a Role in Mathematical Architecture Design (NeurIPS 2023 · similarity 0.91)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.89)
- AtomNAS: Fine-Grained End-to-End Neural Architecture Search (ICLR 2020 · similarity 0.89)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.89)
- einspace: Searching for Neural Architectures from Fundamental Operations (NeurIPS 2024 · similarity 0.88)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · 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.