Research map / Architectures and efficiency / Neural architecture search
PA&DA: Jointly Sampling Path and Data for Consistent 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 |
|---|---|---|---|
| jointly sample paths and training data using normalized gradient-norm importance | SPOS supernet training · NAS-Bench-201, CIFAR-10 | Kendall's Tau: 0.639 ± 0.030 -> 0.713 ± 0.002 | Helped |
| jointly sample paths and training data using normalized gradient-norm importance | SPOS supernet training · NAS-Bench-201, CIFAR-10 | P@Top5%: 0.211 ± 0.168 -> 0.301 ± 0.018 | Helped |
| use PA&DA for architecture search | DARTS search space · CIFAR-10 | Best test accuracy (%): 97.60 -> 97.66 | Helped |
| use PA&DA for architecture search | DARTS search space · CIFAR-10 | Average test accuracy (%): 97.51 ± 0.08 -> 97.52 ± 0.07 | Helped |
| use PA&DA for architecture search | ProxylessNAS search space · ImageNet validation | Top-1 accuracy (%): 77.1 -> 77.3 | Helped |
| use PA&DA for architecture search | ProxylessNAS search space · ImageNet validation | Top-5 accuracy (%): 93.3 -> 93.5 | Helped |
| add data importance sampling (ablation) | SPOS supernet training · NAS-Bench-201, CIFAR-10 | Kendall's Tau: 0.639 ± 0.030 -> 0.644 ± 0.014 | Helped |
| add data importance sampling (ablation) | SPOS supernet training · NAS-Bench-201, CIFAR-10 | P@Top5%: 0.211 ± 0.168 -> 0.225 ± 0.049 | Helped |
| add path importance sampling (ablation) | SPOS supernet training · NAS-Bench-201, CIFAR-10 | Kendall's Tau: 0.639 ± 0.030 -> 0.699 ± 0.004 | Helped |
| add path importance sampling (ablation) | SPOS supernet training · NAS-Bench-201, CIFAR-10 | P@Top5%: 0.211 ± 0.168 -> 0.299 ± 0.008 | Helped |
| use instance-wise rather than class-wise data sampling while increasing τ (ablation) | DA · NAS-Bench-201, CIFAR-10 | Kendall's Tau: 0.637 ± 0.024 -> 0.644 ± 0.014 | Helped |
| increase rather than decrease τ for instance-wise data sampling (ablation) | DA · NAS-Bench-201, CIFAR-10 | Kendall's Tau: 0.643 ± 0.021 -> 0.644 ± 0.014 | No clear effect |
Closest papers by meaning
- GreedyNAS: Towards Fast One-Shot NAS With Greedy Supernet (CVPR 2020 · similarity 0.90)
- K-shot NAS: Learnable Weight-Sharing for NAS with K-shot Supernets (ICML 2021 · similarity 0.89)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.87)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.87)
- SWAP-NAS: Sample-Wise Activation Patterns for Ultra-fast NAS (ICLR 2024 · similarity 0.85)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.85)
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.