Research map / Architectures and efficiency / Neural architecture search
Landmark Regularization: Ranking Guided Super-Net Training in 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 |
|---|---|---|---|
| Add landmark ranking regularization to NAO | NAO · NASBench-101 (CIFAR-10) | S-KdT: 0.329 ± 0.11 -> 0.457 ± 0.03 | Helped |
| Add landmark ranking regularization to GDAS | GDAS · NASBench-201 (CIFAR-10) | S-KdT: 0.691 ± 0.01 -> 0.755 ± 0.01 | Helped |
| Add landmark ranking regularization to GDAS | GDAS · DARTS search space (CIFAR-10) | Mean Acc.: 90.48 ± 2.95 -> 94.32 ± 0.28 | Helped |
| Add landmark ranking regularization to NAO | NAO · ImageNet | Mean Top-1: 67.70 ± 0.43 -> 68.89 ± 0.58 | Helped |
| Add landmark ranking regularization to SPOS | SPOS · ReDWeb validation, V2 Sync | Best Val. loss: 0.0957 ± 0.001 -> 0.0936 ± 0.000 | Helped |
| Use cosine increasing rather than cosine decreasing regularization (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.712 -> 0.803 | Helped |
| Use cosine increasing rather than constant regularization (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.753 -> 0.803 | Helped |
| Set lambda_max to 100 rather than 10 (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.803 -> 0.793 | Hurt |
| Use ten rather than three landmark iterations (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.802 -> 0.811 | Helped |
| Use one rather than three landmark iterations (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.802 -> 0.760 | Hurt |
| Add one sampled landmark pair per training step instead of no pair (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.771 -> 0.803 | Helped |
| Increase sampled landmark pairs from one to twenty per training step (ablation) | SPOS+Landmark Regularization · NASBench-201 (CIFAR-10) | S-KdT: 0.803 -> 0.812 | No clear effect |
Closest papers by meaning
- Distribution Consistent Neural Architecture Search (CVPR 2022 · similarity 0.92)
- AlphaNet: Improved Training of Supernets with Alpha-Divergence (ICML 2021 · similarity 0.91)
- Boosting Order-Preserving and Transferability for Neural Architecture Search: a Joint Architecture Refined Search and Fine-tuning Approach (CVPR 2024 · similarity 0.90)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.89)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.89)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · 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.