Research map / Architectures and efficiency / Neural architecture search
Stabilizing Differentiable Architecture Search via Perturbation-based Regularization
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 |
|---|---|---|---|
| Add random architecture-weight perturbations during training | DARTS · CIFAR-10 | test error (%): 3.00\% to 2.67\% | Helped |
| Add adversarial architecture-weight perturbations during training | DARTS · CIFAR-10 | test error (%): 3.00\% to 2.61\% | Helped |
| Use worst-case perturbations instead of random perturbations | SDARTS · CIFAR-10 | test error (%): $2.67\pm 0.03$ -> $2.61\pm 0.02$ | Helped |
| Add adversarial perturbation regularization | P-DARTS · CIFAR-10 | test error (%): 2.50 -> 2.48 | Helped |
| Add adversarial perturbation regularization | P-DARTS · ImageNet mobile setting | top-1 test error (%): 24.4 -> 24.2 | Helped |
| Add adversarial perturbation regularization | P-DARTS · ImageNet mobile setting | top-5 test error (%): 7.4 -> 7.2 | Helped |
| Add random architecture-weight perturbations during training | DARTS · PTB valid | perplexity: 58.1 -> 58.7 | Mixed |
| Add random architecture-weight perturbations during training | DARTS · PTB test | perplexity: 55.7 -> 56.4 | Hurt |
| Add adversarial architecture-weight perturbations during training | DARTS · PTB valid | perplexity: 58.1 -> 58.3 | Mixed |
| Add adversarial architecture-weight perturbations during training | DARTS · PTB test | perplexity: 55.7 -> 56.1 | Hurt |
| Use adversarial perturbation regularization | DARTS · CIFAR-100 S1 | test error (%): 29.46 -> 22.33 | Helped |
| Use adversarial perturbation regularization | DARTS · CIFAR-100 S2 | test error (%): 26.05 -> 20.56 | Helped |
Closest papers by meaning
- Understanding and Robustifying Differentiable Architecture Search (ICLR 2020 · similarity 0.93)
- b-DARTS: Beta-Decay Regularization for Differentiable Architecture Search (CVPR 2022 · similarity 0.92)
- DARTS-: Robustly Stepping out of Performance Collapse Without Indicators (ICLR 2021 · similarity 0.91)
- BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule (CVPR 2022 · similarity 0.91)
- EG-NAS: Neural Architecture Search with Fast Evolutionary Exploration (AAAI 2024 · similarity 0.88)
- IS-DARTS: Stabilizing DARTS through Precise Measurement on Candidate Importance (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.