Research map / Architectures and efficiency / Neural architecture search
Cyclic Differentiable Architecture Search
Hongyuan Yu, Houwen Peng, Yan Huang, Jianlong Fu, Hao Du, Liang Wang, Haibin Ling
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 joint learning of search and evaluation networks (L_val^{S,E}) (ablation) | CDARTS · ImageNet | top-1 accuracy: 72.8 -> 75.7 | Improved |
| add updating evaluation network weights during joint training (L_val^E) (ablation) | CDARTS · ImageNet | top-1 accuracy: 75.7 -> 76.6 | Improved |
| add l1-norm regularization to alternating search (ablation) | DARTS-style alternating optimization · ImageNet | search stability: 72.8 | Improved |
| use 20-cell instead of 8-cell evaluation network (ablation) | CDARTS · CIFAR10 | accuracy | Improved |
| increase search epochs to 90 (ablation) | CDARTS · CIFAR10 | top-1 accuracy: 97.77 | Improved |
| scale up to big model (channels 36 to 50) | CDARTS · CIFAR10 | test accuracy: 97.52 -> 98.32 | Improved |
| scale up ImageNet big model (channels 48 to 96, 320x320) | CDARTS · ImageNet | top-1 accuracy: 81.12 | Improved |
| CDARTS vs DARTS | CDARTS · ImageNet | top-1 accuracy: 73.3 -> 76.3 | Improved |
| CDARTS vs DARTS V1 on stability | CDARTS · CIFAR10 five runs | mean accuracy: 96.93±0.13 -> 97.52±0.04 | Improved |
| rank correlation of architecture weights | CDARTS vs SPOS · NAS ranking | Kendall tau: 0.19 -> 0.49 | Improved |
| CDARTS vs DARTS in NATS-Bench | CDARTS · NATS-Bench CIFAR10 test | accuracy: 59.84 -> 94.02 | Improved |
| introspective distillation vs DNA-b teacher | CDARTS-b · ImageNet chain-structured | top-1 accuracy: 77.5% -> 78.2% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DARTSV1 — DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021) · better on 5 of 5 shared results
- FairDARTS — Fair DARTS: Eliminating Unfair Advantages in Differentiable Architecture Search (ECCV 2020) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Improving the Consistency, Efficiency, and Flexibility of Differentiable Neural Architecture Search (CVPR 2021 · similarity 0.92)
- Partially-Connected Neural Architecture Search for Reduced Computational Redundancy (TPAMI 2021 · similarity 0.91)
- iDARTS: Differentiable Architecture Search with Stochastic Implicit Gradients (ICML 2021 · similarity 0.91)
- DARTS: Differentiable Architecture Search (ICLR 2019 · similarity 0.91)
- Understanding and Robustifying Differentiable Architecture Search (ICLR 2020 · similarity 0.90)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · 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.
